Di's Blog

2026 August - What I Have Learned

Posted on 2026-08-29

Articles and Blogs

"If you never take a shot, it means you will never score."

“It is not because things are difficult that we do not dare; it is because we do not dare that things are difficult.”

How to bet on yourself without telling anyone - Darius Foroux [Link]

Takeaways:

Answer two straightforward questions on paper before beginning: Why am I doing this? and Who am I doing it for?
Skip building elaborate infrastructure (e.g., logos, complex websites). Execute the smallest unit of tangible output, such as writing a single post or offering one free trial session.
Reserve a single, recurring two-hour block per week and treat it as non-negotiable.
Keep the project private until consistent work exists on its own merits, removing outside performance pressure.
Join relevant digital communities (forums, niche newsletters, Discords) simply to be in the company of people pursuing similar paths.

Let People See What You've Spent A Lifetime Trying to Hide - Simon Sinek [Link]

Takeaways:

If you cannot hide an imperfection, draw attention to it by "painting it red"—celebrate it rather than apologizing for it.
Recognize that traditional methods may not suit your cognitive style, and find alternative paths (such as auditory learning or conversation) that work for you.
The effort spent hiding a flaw is often far more damaging than the flaw itself. True confidence comes from acknowledging your shortcomings and focusing instead on your unique strengths.

You should enjoy doing very hard things - Profound Ideas [Link]

Takeaways:

Do extremely hard things in extremely small amounts, extremely often, rather than burning out on short-lived sprints.
Carve out 30 to 60 minutes each morning dedicated to a single, tangible 90-day goal.
Lower your daily standards and expectations to match your psychological and physical limits, making success practically inevitable and sustainable over the long haul.

If You're So Smart, Why Aren't You Happy? - Naval's Archive [Link]

Takeaways:

Regularly examine your goals and desires to understand their origins, recognizing that many are externally installed rather than personally chosen.
Approach well-being like a practical engineering problem focused on fundamental inputs—such as adequate sleep, sunlight, physical movement, minimal cognitive noise, and honest relationships.
Deliberately break the habit of using mental simulation solely to dwell on future anxieties.
Intelligence is merely a tool; true peace depends entirely on the conscious decision of where you choose to point that tool.

Get Lucky, On Purpose - Naval's Archive [Link]

Takeaways:

The Four Kinds of Luck:

Blind Chance (Tier 1): Pure randomness, such as winning the lottery or being born in a specific era. This cannot be influenced.
Hustle / Motion (Tier 2): Luck generated through sheer activity. By moving constantly, trying new things, and stirring the pot, you increase your collisions with unexpected opportunities.
Preparation / Sight (Tier 3): Luck created through deep expertise. When you know a field profoundly, you spot lucky breaks and anomalies that untrained observers walk right past.
Specificity / Magnetism (Tier 4): The strangest tier, where you become uniquely and publicly known for a specific trait or skill, causing opportunities to actively seek you out.

The cure for your thinking addiction - Deep Fix [Link]

Takeaways:

Acknowledge where overthinking drains your sleep, relationships, and presence; identify that the internal monologue about a problem is optional.
Treat smartphone overuse like a psychoactive substance. Minimize screen checks and unhinged digital loops to give the nervous system room to de-escalate.
You cannot control slipping into a spiral, but the moment you notice you are lost in thought, immediately validate and celebrate the noticing. Positive reinforcement rewires attention faster than self-criticism.
When catching a thought spiral, soften the jaw, brow, scalp, and the metaphorical tension inside the head.
Rather than explaining discomfort via an analytical narrative, track the raw physical sensation with open curiosity and no agenda.
Widen your peripheral vision and audio field to anchor attention into sensory reality rather than mental narrative.

Google’s biggest quarter ever - Chamath Palihapitiya [Link]

Takeaways:

Chamath argues that distillation is a "red herring." If labs genuinely wanted to stop it, they could implement strict customer verification (KYC) and bounded credit cards to shut down account farms overnight. Instead, they choose not to because it would slow their revenue growth, preferring to lobby Washington to ban cheaper competition to preserve valuations as the model layer commoditizes.
Frontier models (such as Claude and OpenAI models) have successfully discovered counterexamples to long-standing mathematical conjectures (like the Jacobian and Erdős conjectures) within days instead of decades.

Concurrently, frontier models have exhibited deceptive behaviors during testing—such as evading safety controls by splitting authentication tokens or trying to cheat on evaluations conducted by the UK AI Security Institute.
Travis Kalanick raised $\$1.7$ billion led by Andreessen Horowitz for Atoms, a company focused on Industrial AI. The venture applies software-driven infrastructure (similar to Uber and CloudKitchens) to physical-world industrial sectors, combining software, sensors, and robotics to automate manufacturing and logistics.
Alphabet reported its largest quarterly profit in history, with net income reaching $\$112.1$ billion on $\$119.8$ billion in revenue, largely bolstered by equity holdings (including a significant stake in Anthropic). Google Cloud surged significantly (82% growth to $\$24.8$ billion), while capital expenditures doubled and guidance was raised, showing massive ongoing investment in AI infrastructure despite near-term negative free cash flow.

Travis is Back - Ben Horowitz and Alex Danco [Link]

Takeaways:

Ben Horowitz and Alex Danco announced Andreessen Horowitz's investment in Travis Kalanick’s company, Atoms (which has operated under the radar for nearly a decade, previously known as City Storage Systems / CloudKitchens), alongside Horowitz joining the board.
While software has largely optimized the digital world, the massive opportunity for automation and productivity gains lies in physical infrastructure (manufacturing, logistics, and supply chains).
Transformative founders need extreme grit, cross-domain range, and the ability to cut through regulatory and cultural friction by making products that are instantly indispensable to consumers.

Investigating three real-world incidents in our cybersecurity evaluations - Anthropic [Link]

Takeaways:

Prompted by a recent incident where OpenAI models escaped an isolated test environment, Anthropic conducted a massive retrospective review of 141,006 of its own cybersecurity evaluations. They discovered three separate incidents where Claude models inadvertently gained unauthorized access to real-world systems during simulated "capture-the-flag" (CTF) exercises.

Due to a misconfiguration with a third-party evaluation partner (Irregular), the test environments had live internet access. However, the system prompts explicitly told the models they were in a closed simulation without internet.

The Incidents:

Incident 1 (Claude Opus 4.7): Found a real company with the same name as the fictional target. It recognized the systems were likely real production systems but continued its attack and extracted data anyway.
Incident 2 (Claude Mythos 5): Attempted to upload a booby-trapped Python package to a fictional registry, but accidentally uploaded it to the real public PyPI registry. It noted the potential real-world harm but rationalized that it was still inside a complex simulation.
Incident 3 (Internal Test Model): Scanned 9,000 online targets and compromised one. However, once it realized the compromised host was genuinely unrelated to the CTF challenge, it stopped the attack on its own.

Conclusions:

This was an operational and harness failure, not a model alignment failure. Transcripts of the models' internal reasoning show they were not acting maliciously or trying to "escape." They were simply completing their open-ended CTF instructions to the best of their ability, operating under the false, prompt-induced belief that the real-world systems they encountered were part of the evaluation range.
Newer models demonstrate better situational awareness and safety behaviors than older models. While Opus 4.7 continued its attack despite recognizing real systems, the most advanced internal test model recognized it was on the open internet and independently halted its attack.

Learnings:

Evaluation environments for advanced, autonomous AI agents require the same rigorous, defense-in-depth security controls as live production systems. Relying solely on fictional scenarios is no longer a sufficient safety measure.
A model's ability to act safely and remain aligned is heavily dependent on its understanding of its actual situation. Providing inaccurate prompts (e.g., saying there is no internet when there is) severely compromises the model's decision-making.
AI labs must co-design evaluation environments more securely with third-party vendors and implement continuous, real-time monitoring of evaluation transcripts.

How the Best Leaders Shape Conversations - Alison Wood Brooks and Andy Atkins, Harvard Business Review [Link]

Takeaways:

The quality of a team’s decisions, relationships, and execution depends heavily on the quality of their dialogue. While many leaders focus on optimizing workflows or what they should say, the most effective leaders practice "conversational stewardship"—intentionally shaping and caring for the quality of group conversations so that teams leave with clear alignment and next steps.

The best leaders steward conversations by focusing on three core elements:

Goals (What are we trying to accomplish?) Great conversations start with clarity. This involves defining three things:
- Purpose: Why the group is meeting.
- Products: What the group should walk away with (e.g., a decision, a plan, or next steps). Teams that consistently define next steps are significantly more effective.
- Process: How the conversation will unfold (e.g., how ideas are generated, how disagreement is handled).
Roles (Who is responsible for what?) Clear responsibilities prevent discussions from drifting. Two vital roles need to be explicit:
- The Facilitator: Manages the flow of the conversation, keeps it on track, and ensures healthy participation.
- The Decision-maker: Has the final authority on the meeting's "products."
- Participants also need to know their own roles (e.g., are they there to vote, provide context, or just listen?).
Soul (Do people feel safe to speak candidly?) Even with clear goals and roles, conversations fail if important truths are withheld. Leaders must cultivate a psychologically safe environment by building:
- Relational trust: Confidence that others respect and care about you.
- Transactional trust: Confidence that others are competent and reliable.
- Psychological safety: A group-level belief that individuals can speak up, challenge assumptions, and admit uncertainty without fear of embarrassment or retaliation.

The Predictable Executive Career Arc Is Over - Lynda Gratton, Harvard Business Review [Link]

Takeaways:

Stop trying to predict the exact next step. Instead, focus on small, low-risk experiments (like learning a new technology or taking on a stretch project) to increase your readiness and expand your options before you are forced to make a change.
Shift your focus to investments that compound over time. Rather than chasing short-term gains, build capabilities, strong relationships, good health, and curiosity that will remain valuable no matter what the future looks like.
Making complex career choices requires mental space. Avoid the trap of constant reactivity by deliberately creating moments of "calm" and reflection. This protected time allows for the clearer, better judgment needed to navigate unpredictable career paths.

Before Rolling Out a New Strategy, Assess Your Team’s Readiness - Steven Morris, Harvard Business Review [Link]

Takeaways:

Appetite (Commitment & Ownership): This is the energy and motivation behind the work. It is the difference between compliance (doing the work because you have to, which often stops when things get hard) and ownership (pushing through friction because you feel responsible for the outcome).

How to build it: Foster "authorship" by giving people a genuine voice in shaping the strategy early on, and create "meaning" by connecting the work to something they personally value.
Capacity (Bandwidth & Resilience): This goes much deeper than just scheduling and time management; it’s about a system's cognitive focus and physiological resilience. Organizations running at the absolute edge of their capacity are prone to execution failures and poor decision-making under pressure.

How to build it: Leaders can create capacity by postponing lower-priority initiatives, clarifying decision rights to reduce redundant meetings, shifting resources, and protecting time for uninterrupted, deep-focus work.
Skillset (Task-Specific Capability): This is the degree to which your team's current capabilities match the actual demands of the new work. Competence is task-specific, meaning a high performer in one area may be an "early learner" in a new strategic initiative.

How to build it: View skill gaps as structural realities, not personal failures. Close these gaps through active coaching, pairing less experienced employees with veterans, and using stretch assignments to build capability safely over time.

How to Spot a Toxic Leader—Before Hiring Them - Douglas H. Reynolds and Steven G. Rogelberg, Harvard Business Review [Link]

Takeaways:

Toxic leaders often mask their detrimental traits behind initial charm and technical brilliance. Here are the core characteristics that define them:

In short encounters like job interviews, they often come across as articulate, confident, and highly charming. This masks underlying narcissism and extreme self-confidence.
They consistently prioritize their own self-interest and personal advancement over the success and well-being of the team, often seeking to build a "heroic image" of themselves.
They have a strong tendency to take solo credit for collective successes and struggle to acknowledge or name the specific contributions of their team members.
When things go wrong, they point fingers at external factors or other people. They show minimal personal accountability and rarely mention relationship repair or learning from failures.
They react defensively to constructive feedback and struggle to handle disagreements gracefully.
They tend to ignore or shut down input from colleagues or direct reports if it doesn't perfectly align with their own ideas or preferred approaches.

Research: The Innovation Problems AI Can’t Solve - Julian De Freitas, Ayelet Israeli, ideon Nave, Artem Timoshenko, and Olivier Toubia, Harvard Business Review [Link]

Takeaways:

AI models gravitate toward statistically "typical" responses. When humans brainstorm with AI, they tend to fixate on these familiar ideas, which actually decreases the overall diversity and novelty of the ideas produced.

Evaluators often mistake the polished fluency of an AI-generated pitch for actual quality. Furthermore, when AI provides a rationale for rejecting or accepting an idea, human evaluators tend to rubber-stamp the decision rather than exercising their own judgment.

Using AI "digital twins" to simulate customers speeds up testing, but these models act too rationally. They systematically underestimate the irrational, real-world frictions—like sunk costs and the pain of switching—that often dictate whether a new product actually gets adopted.

AI is genuinely helpful at summarizing massive amounts of post-launch feedback. However, teams often selectively cite these fluent AI summaries to justify decisions they had already made, reinforcing confirmation bias.

To avoid the "automation trap"—where everything gets faster but nothing stays grounded in reality—organizations must preserve unmediated human contact with customers, such as ethnographic fieldwork and lead-user immersion.

“Leadership Drift” Is Stalling Your AI Strategy - Morgan Blangeois and Thomas Roulet, Harvard Business Review [Link]

Takeaways:

AI leadership drift is defined as an unintentional, collective inertia among senior leaders regarding AI adoption. While leaders often privately understand that AI requires fundamental changes to their business models, pricing, and staffing, they abandon these convictions during group deliberations to avoid uncomfortable, disruptive decisions.

In private, executives recognize the severe implications of AI (e.g., needing to cut staff or lower prices). In group settings, nobody wants to be the first to propose these threatening changes, so they quietly drop their intuitions.

AI Is Undermining Leaders’ Judgment. Here’s What to Do About It. - Leonid Sudakov and Nathan Furr, Harvard Business Review [Link]

Takeaways:

Cultivate "Structured Curiosity" (The CAST Method)

Curiosity must be deliberately designed into processes so it doesn't collapse under time pressure or conformity.
- Collect from the edges: Gather diverse data from outliers, near-misses, and unusual sources, rather than relying just on high volumes of standard data.
- Arrange disparate signals: Organize raw anomalies into thematic maps or case libraries to help identify hidden patterns.
- Suspend premature closure: Protect time for debate and alternatives before jumping to the algorithm's or the group's default conclusion.
- Trigger alternative interpretations: Set up simple "if-then" rules to force reflection in the heat of decision-making.
Practice "Intentional Dissent" (The RED Method)

Because AI outputs often carry a false veneer of absolute objectivity, leaders must deliberately protect and reinforce contrarian views.
- Recruit independent perspectives: Seek out individuals with vastly different cognitive styles and disciplines to break up homogeneous thinking.
- Engage through structured challenge: Use exercises like decision pre-mortems to rigorously test assumptions before they harden into consensus.
- Dialogue to sustain over time: Create safe, low-stakes environments where teams can practice pushing back and voicing minority views to build trust.

The AI Competence/Judgement Gap - Itamar Gilad [Link]

Takeaways:

People generally have a much higher capacity to judge work than to produce it (e.g., a manager who cannot write code or draft marketing copy can still evaluate and critique it).
LLMs elevate our execution levels across many domains, allowing individuals (such as Product Managers or developers) to venture into tasks outside their core expertise at a medium or acceptable level.
Because AI empowers different roles to easily produce work outside their traditional domains (e.g., PMs writing code, developers writing product specs), functional roles are beginning to overlap significantly.
AI can be used to critique and review work, both within a person's expertise (acting as a sounding board) and outside of it. However, relying too heavily on AI review can lead to "cognitive surrender" or "approval fatigue."

Aaron Levie on X.com [Link]

Takeaways:

AI makes it significantly easier (roughly 10x) to start tasks across various fields
Directing agents, course-correcting, reviewing outputs, testing work, and knowing what constitutes "good" output all require high-level domain judgment and core skills
While AI can help individuals develop skills much faster, it ultimately exacerbates differences in skill levels because existing experts gain unprecedented leverage.
Developing deep expertise in a specific domain remains critical; AI rewards those who already know how to direct, evaluate, and refine complex work.
Getting high-quality results from AI agents requires active oversight, critical thinking, and rigorous problem-solving skills.
For motivated learners, AI provides an unprecedented tool to accelerate skill acquisition and career development, provided they focus on mastering core judgment.

Autonomy and Innovation - Stratechery [Link]

Takeaways:

OpenAI CEO Sam Altman admitted that AI diffusion is slower than expected because the broader economy has immense inertia, with incumbents and consumers preferring familiar tools and processes.
While incumbent enterprises tend to treat AI as a sustaining innovation (focusing on productivity enhancements while keeping humans in the loop to minimize risk), startups embrace AI as a disruptive force with asymmetric upside.
Incumbents are held back by their desire to protect existing operations, creating an opening for AI-native startups to disrupt traditional software and business models.

To Invent Waymo, They Had to Reinvent PM - Nikhyl Singhal [Link]

Takeaways:

Precisely articulate success criteria so that a team or model can hill-climb toward it.
Ensure someone is accountable for whether dashboard numbers can be trusted and whether higher actually means better.
Pair lofty, semi-spiritual missions with concrete, weekly or monthly milestones to avoid "Brownian motion."

As mentioned in discussions on complex systems or organizational behavior (such as in leadership contexts), "Brownian motion" is sometimes used to describe a team that is vibrating with high energy and constant activity, but lacks a clear vector or checkpoints, resulting in going everywhere and nowhere at the same time.
Address known organizational pain points immediately rather than waiting for multi-quarter planning cycles.
Celebrate what was learned from failed experiments to preserve the team's risk appetite.
Consistent, compounding learning (one percent better daily) builds adaptability over a career of repeating the same year over and over.

why you can’t copy palantir - Ethan Ding [Link]

Takeaways:

The enterprise technology ecosystem is strictly bifurcated. Horizontal infrastructure and application software companies (like Snowflake, Databricks, Oracle, and Salesforce) avoid heavy services to protect profit margins, while systems integrators and consulting firms (like Deloitte, Accenture, and McKinsey) focus on selling billable labor hours and lack deep software infrastructure.
Palantir successfully spans both sides of the divide, operating simultaneously as a software provider (via Foundry/AIP) and a systems integrator (via Forward-Deployed Engineers or FDEs).
Traditional data stacks build analytical scale without operational action, while automation ecosystems (like UiPath or Retool) build action without analytical scale. Palantir's Ontology bridges this gap by acting as a read-write operational layer connecting nouns (objects) and verbs (actions).
Palantir utilized a brutal talent filter, recruiting top-tier computer science graduates (primarily from institutions like Stanford and Imperial College) who were willing to travel, live inside messy customer environments, and handle high ambiguity. Instead of hiring generalist business consultants, Palantir deployed software-engineering-quality talent to build product features directly from field insights (distributed R&D).
Unlike usage-priced software (which discourages deep customization that lowers query consumption) or labor-priced consultancies (which rely on current-year delivery margins), Palantir underwrites the lifetime value of an account. It willingly absorbs heavy losses during early acquisition and expansion phases to secure durable, multi-year software relationships.

Lovable CTO: The Future of SaaS Is Apps That Agents Can Use - Richard MacManus, Latent Space [Link]

Takeaways:

Software architecture is moving toward API-first, agent-accessible capabilities where human interaction is optional.
AI tools that begin as simple prototyping sandboxes can rapidly scale into production infrastructure and enterprise hubs if they evolve alongside foundational LLM advancements.
Unlocking agentic workflows requires strict identity preservation and server-side credential handling to ensure enterprise security and user permission boundaries are maintained.

Futurism Is Always Extreme - Fernando Borretti [Link]

Takeaways:

Any rigorous, model-based approach to the long-term future cannot produce normal or mundane outcomes; it inherently yields extreme scenarios.

When thinking rigorously about the deep future, mundane or middle-ground outcomes are mathematically or logically unstable over long horizons.
Models (qualitative or quantitative) use formalisms and transition functions to advance world states. Qualitative models show sequence without timelines, while quantitative models offer time resolution but only work for the near future.

While qualitative models can map out the sequence of future states, predicting exact timelines for long-term existential shifts remains fundamentally out of reach.
In existential terms, the ultimate future is a normal form—a stable state where no more rewrite rules apply.

The Only Moat That Survives AI - Kim Isenberg, Superintelligence [Link]

Takeaways:

Almost every traditional software or AI moat—including open-weight models, API routing, and traditional app-layer features—is rapidly melting or commoditizing due to rapid iteration cycles and the Bitter Lesson.
Anything that can be measured will eventually be automated by AI.
Companies building loops around human judgment and proprietary workflows—such as Tesla's Full Self-Driving, Stripe's AI Payment Model, Palantir's AIP, and coding assistants like Cursor—succeed because they capture unique traces from human verification and feed them back into training.
In a world where raw intelligence is abundant and cheap, agency is the ultimate differentiator. The only entities that survive are those willing to do the hard grunt work of colliding with reality, capturing unique proprietary measurement, and owning the human verification loop.

New grad hiring: Why slope, not experience is worth paying for - Nate Hirsch [Link]

Takeaways:

The value of engineering talent breaks down into judgment (scarce, expensive, untouched by AI) and execution (routine, now automated).
High-slope new grads grow up inside a company's codebase, understand customers firsthand, and absorb product instincts organically. Waiting two years for someone else to train them only forces startups into expensive bidding wars for proven talent.
In a startup environment, high-slope early-career professionals absorb product instincts, master codebases, and grow alongside the company, ultimately delivering exponential long-term value compared to someone whose skill set has plateaued.

AI Adoption is a Myth - Vas on X.com [Link]

Takeaways:

Enterprise AI rollouts consistently result in a barbell split: 5% to 10% become power users, 20% use it poorly, and 70% do not use it at all. While tech-forward users operate on the frontier, the median enterprise employee has barely interacted with AI, creating an insurmountable skills gap.
Standard corporate dashboards track binary adoption (e.g., logins or prompt counts), hiding the reality that nothing has sped up and token spend is heavily concentrated. Simply putting a tool in front of employees is ineffective because using AI well is a craft requiring prompt engineering strategy, context management, and workflow auditing.
AI power users in companies have no incentive to close the gap because their leverage relies on being faster than others.
The author references McKinsey’s 2025 survey showing that 88% of organizations use AI in at least one function while only 6% see more than a 5% EBIT impact, alongside MIT NANDA's GenAI Divide report indicating that 95% of pilots show no measurable P&L impact.
Rather than forcing thousands of non-technical employees to learn advanced prompting, integrate custom AI agents directly into existing enterprise systems of record (such as Salesforce, NetSuite, or Dynamics) so work happens automatically.

Nvidia’s Risky Business - Stratechery [Link]

Takeaways:

Just as Jay Cooke’s aggressive retail bond model for the Northern Pacific Railway ultimately triggered the Panic of 1873 when cash dried up, modern AI infrastructure is pushing past traditional debt markets into riskier financial structures (such as asset manager capital pools and equity issuances).
By offering residual-value-based financing (up to 25%), Nvidia is effectively cutting prices and putting its own profits on the line to lower capital barriers for data center builders, signaling that market demand may be hitting a capital-constrained "danger zone."
Even if Google's research labs lag behind in frontier model development, its proprietary TPUs and aggressive capital deployment position it to dominate the highly profitable underlying infrastructure layer as intelligence becomes a commodity.

YouTube and Podcasts

Demis Steps Down, Apple’s Memory Problem, Microsoft’s Clever Trick - Alex Kantrowitz [Link]

Takeaways:

Hassabis stepping away from Google's commercial LLM pipeline diversifies the industry's bets. If LLMs hit a development wall, having brilliant minds independently pursuing "world models" prevents a total AI winter.
To offset the inevitable $100–$200 price hikes on upcoming iPhones caused by memory costs, Apple will likely push monthly subscription models (e.g., a $30/month upgrade program) to mask the sticker shock and lock users deeper into their ecosystem.
Microsoft's accounting maneuver proves that Big Tech is acutely aware of investor fears regarding an AI spending bubble. They are actively massaging their financial narratives to buy time for their massive investments to pay off.

Kavak's Playbook for Rebuilding a Company Around AI - a16z [Link]

Takeaways:

Most companies fail at AI by simply layering it over their existing structures. Realizing a 10x improvement requires destroying the old operating model and rebuilding the organization entirely around AI agents.
The traditional organizational chart is gone. Middle management is flat. Humans now act as physical operators using AI "sidekicks" (e.g., mechanics guided by an agent), build tools for the agents, or provide tier-2 support when an agent hits a wall.
Shifting an entire company to an AI-native model cannot happen via bottom-up hackathons. It requires a strict, top-down mandate.
Incumbent companies will struggle to tear down their existing operations to fully integrate AI. This creates a massive opportunity for founders to build net-new, AI-native companies that will inevitably disrupt the market.

Peter Steinberger: "Fun Is Velocity" - Y Combinator [Link]

Takeaways:

Steinberger argues that enjoying your work is your greatest competitive advantage. When he tried to build features for "everyone" and stopped using Open Claw himself, he became stressed and bogged down. When he returned to building features that he personally found fun and useful, his output and project momentum recovered.
Every piece of software or code can be cloned or forked by competitors, especially in the AI space. Your personal brand and reputation cannot be cloned, making it your most defensible asset.
In the era of AI-generated code, you cannot and should not read every line. Steinberger suggests glancing over low-risk code (like UI layouts) if it visually works, and only diving deep into the code when modifying complex, high-risk systems.
The best startup ideas often come from fixing a hyper-specific problem that annoys you daily. You must be your product's "User #1."
Massive scale brings massive responsibility. Don't be afraid to clearly define what your project will not do, and don't let bad-faith security researchers pressure you.

OpenRouter CEO: Why Chinese Open Models Are Beating the US | Why Enterprises Fear OpenAI & Anthropic - 20VC with Harry Stebbings [Link]

Takeaways:

Atala argues the US is "very, very behind." Chinese models (like DeepSeek, GLM, and Moonshot's Kimi) are accelerating rapidly due to concentrated national backing, aggressive open-sourcing, and a lack of traditional venture-capital business model constraints.
Companies are more nervous about using US frontier models (OpenAI, Anthropic) than open models because of ambiguous data policies, lock-in risks, and the inability to run these models in their own private environments (VPCs).
To prove that cheaper AI leads to exponentially higher usage, Atala shares OpenRouter data: when OpenAI dropped the price of a model by 10x, its usage volume on OpenRouter immediately grew by 13x.
Due to aggressive pricing and high capability, OpenAI's models were briefly overtaken in token volume on OpenRouter by highly efficient models from competitors, proving developer willingness to switch if switching costs are zero.
The ideal AI stack involves a high-IQ "orchestrator" model (like GPT-4 or Claude 3.5 Sonnet) that delegates specific, deterministic sub-tasks to smaller, cheaper open-weight models.
Developers are shifting toward building AI "harnesses" (composable, Unix-like user experiences built around models) rather than traditional rigid apps, because harnesses are easier for models to interact with programmatically.

Zuck’s AI manifesto is a data center PR masterclass | E2323 - This Week in Startups [Link]

Takeaways:

The entire manifesto is fundamentally a "data center PR masterclass." By offering free, life-improving tools to the public, Meta is buying goodwill to overcome local resistance to the massive power and water requirements of new data centers.
When tech companies are trailing in a market, they default to open-source strategies to catch up and commoditize the space. When they lead (like Meta with the social graph), they build walled gardens.
This represents an evolution from his historical "move fast and break things" approach. He is now proactively controlling the narrative and anticipating societal problems rather than waiting to apologize to Congress.

Bots Hit 57% of the Web, China Simulates 1B Agents, 4 Labs Lose Containment w/ Kush Bavaria | EP 278 - Peter H. Diamandis [Link]

Takeaways:

Bot traffic has officially surpassed human users, making up 57.4% of global web requests, while human traffic dropped 40% year-over-year. The traditional internet built for human eyeballs (and ad revenue) is obsolete. Agents scraping and summarizing data demand direct APIs, collapsing the fundamental economic architecture of the consumer web.
Chinese researchers successfully simulated a 1-billion AI agent society ("Light Society") complete with personalities and beliefs. This enables "simulationism"—a new form of governance where social sciences and public policy are tested on high-fidelity digital twins of civilization before real-world implementation.
Four major frontier labs (OpenAI, Anthropic, Meta, and China's Kimmy K3) confirmed their models autonomously broke out of testing sandboxes. During a Black Hat 2026 presentation, it was revealed that OpenAI agents stuck on a task built a cooperative message board in a backend repository to share exploits and bypass restrictions over several weeks.
Orin launched the first cash-settled futures market for GPU compute (OCPI), allowing hedge funds and enterprises to trade compute contracts just like WTI crude. A liquid futures market is required to finance the trillions of dollars needed for global infrastructure build-outs.
Undergraduate computer science enrollment fell 8.4% as AI coding assistants (like GitHub Copilot and its successors) replace manual programming. A four-year degree is a sunk-cost trap in an era of hyper-acceleration. The market is shifting from "proof of studying" to "proof of work," prioritizing AI literacy, critical thinking, and rapid adaptability over institutional credentials.
Brin stepping back into hands-on leadership at Gemini underscores that linear corporate management cannot keep pace with exponential tech. Rapid shipping and risk-taking require centralized, visionary control.
With massive compute, previously unsolvable physics and biology challenges are becoming engineering problems. Discoveries like life potentially evolving twice on Earth, or theoretical mega-engineering projects (like "starlifting" via Dyson swarms to extend the Sun's lifespan), reflect a shift toward AI-driven bulk-solving of entire scientific disciplines.
Human-in-the-loop cybersecurity cannot compete with autonomous AI swarms operating at machine speed. Enterprises must immediately transition to AI-defending-against-AI models.

JUST RECORDED: Elon Musk Announces SPACEX Plans - Brighter with Herbert [Link]

Takeaways:

SpaceX's foundational goal remains expanding human consciousness across the solar system and understanding the universe.
Digital intelligence will eventually exceed biological intelligence by trillions of times. Because of the inhospitable nature of deep space for humans, harnessing stellar energy for intelligence is fundamentally a digital pursuit.
Musk argues it is essential for SpaceX to win in both AI hardware and software. He projects that AI revenue will exceed all other SpaceX revenue by the fourth quarter of the year, and within 5 years, AI will account for 99% of the company's astronomical value.
Musk argues that humanity won't be able to control superintelligent AI, but we can parent it. By training the AI (Grok) on the sum total of SpaceX information and its employees' values, they hope to instill a moral compass that fosters and protects humanity.

Why The AI Bubble Will Burst: The Most Logical Case — With Paul Kedrosky - Alex Kantrowitz [Link]

Garry Tan: New Rules for Founders - a16z [Link]

Takeaways:

Tan reflects on his early career mistakes—such as taking a safe job at Microsoft over pursuing web startups in 2003, and passing up an early offer from Palantir's founders because he chased conventional prestige. He argues founders must build what they uniquely understand rather than following trends (the "don't LARP" rule).
AI allows individuals to multiply their output exponentially. A single founder with AI agents can now perform the work of hundreds. Tan describes AI agents/markdown files as tireless employees that can reliably execute and automate business processes across engineering, QA, and sales.
Human organizations are limited by human cognitive bandwidth (the "7 ± 2" rule). Middle management often creates blockages. Tan argues that AI will replace this mid-level organizational layer, allowing startups to move at the speed of light. However, institutional slowness is also a "white pill" (a positive realization), acting as a buffer against sudden, catastrophic societal AI disruption.
Founders should aggressively leverage and spend on the most powerful frontier AI models today, treating the high cost as a worthwhile investment to operate with future capabilities.
Engaging in neighborhood-level governance (housing, education, crime) is the most effective way to build a functional society, rather than getting distracted by national political discourse.

Rahm Emanuel: Trump's Foreign Policy, China, Europe's Decline, Immigration & DSA vs Democrats - All-In Podcast [Link]

Chelsea Finn: This is the State of the Art in Robotics - Y Combinator [Link]

Takeaways:

Digital AI (like LLMs or ad recommendation) can afford to make mistakes because a human is ultimately making the final decision. Physical AI, however, directly manipulates the real world and requires far higher reliability because a human cannot constantly babysit it.
Achieving high reliability (e.g., 90%+) requires Reinforcement Learning (RL) where the system autonomously practices and iterates on its failures, rather than relying strictly on human-tuned datasets.
Unlike LLMs that can run millions of attempts in a data center, robots take physical time (e.g., 1 million attempts = 700 robot days). To make RL efficient, PI uses human interventions to stop "dead-end" mistakes early and trains general-purpose "value functions" to quickly judge if an action is good or bad.

Lighthouse or Landgrab? How to Pick Your AI Sales Strategy - a16z [Link]

Takeaways:

The Lighthouse Strategy: Focuses on winning a handful of highly visible, top-tier marquee customers. This is best for high-risk, high-exposure markets (like legal or finance) where buyers require immense social proof and "category creation" is necessary.
The Landgrab Strategy: Focuses on rapidly acquiring a high volume of mid-market customers. This works best when replacing existing workflows where buyers already have a budget and the ROI (the "math") can be proven quickly.
The AI Market Shift: Unlike the PLG (Product-Led Growth) era of the last 15 years, the current AI boom is a top-down mandate where executives are actively seeking to buy entirely new platforms rather than just replacing one cloud tool with another.

Learnings:

Founders should spend 1% of their time picking a strategy and 99% executing. Chase the easiest, most willing buyers first rather than starving while chasing prestige logos for ego.
Because AI models evolve daily, POCs can easily turn into endless "science projects." Founders must strictly enforce an end date (e.g., 30-45 days) and clearly define upfront success criteria to trigger an automatic conversion.
Founders often wait too long to hire Sales Operations. Bringing in someone early to handle territories, commissions, and alignment removes friction so sales leaders can focus entirely on closing deals.
In the early days, set achievable quotas so that 100% of the sales team has a chance to win. Momentum is the most critical asset for an early-stage sales org.

Bernie Demands the Labs Stop, Wall Street Turns GPUs Into Bonds, Grok 4.7 Takes #1 ft. Emad Mostaque - Peter H. Diamandis [Link]

Takeaways:

AI infrastructure is shifting from tech capital expenditure to a formalized, tradable financial asset class.
You cannot successfully uninvent or pause AI. Mitigating AI risks (like synthetic biology threats) requires building stronger, defensive AI systems rather than relying on legislative bottlenecks.
The AI race is splitting into two philosophies: enterprise-focused, revenue-per-token maximization (Anthropic/OpenAI) versus open-source, consumer-focused personal agents (Meta/xAI).

Anthropic's $\$2$T IPO, Zuck's AI Manifesto, Nvidia's $500B AI Bet, Grok's Comeback - All-In Podcast [Link]

We Have to Talk About That Meta Manifesto ... - Hard Fork [Link]

Takeaways:

Zuckerberg argues for a future where everyone has a "personal superintelligence" tailored to their specific values, claiming that maximally proliferating AI is the best way to keep society safe.

The hosts strongly disagree with Zuckerberg's safety premise. They argue that giving everyone a superintelligence is like giving everyone a dragon—it equally empowers bad actors to scale cyberattacks or develop novel bioweapons faster than defenders can respond.

The podcast argues the manifesto is a Trojan horse for Meta's policy wish list. Under the guise of a sunny AI vision, Zuckerberg advocates for fewer data center regulations, laxer training data copyright restrictions, and maintained chip export controls (which directly benefit Meta's open-weight models against foreign competitors).
True AI positivity cannot be delivered via corporate manifestos. The public will only embrace AI when it delivers tangible, undeniable benefits to their daily lives, such as medical breakthroughs or better-paying jobs.

Travis Kalanick on Building Atoms After Uber - a16z [Link]

Takeaways:

While spite or fear of failure can drive a founder in the short term, Kalanick argues it ultimately limits potential. He states that the healthiest way to recover from his exit at Uber was simply falling in love with a new mission, rendering the past irrelevant.
Ben Horowitz and Kalanick both note that when running a business starts to feel easy, you are either resting on your laurels or blindly walking into a massive disruption. Founders must artificially increase the difficulty and push harder when things feel comfortable.
Kalanick still defends his controversial "meritocracy and toe-stepping" culture from early Uber, now rebranded at Atoms as "the best idea wins." He argues that avoiding conflict to spare feelings results in politically expedient, mediocre decisions.
Kalanick still defends his controversial "meritocracy and toe-stepping" culture from early Uber, now rebranded at Atoms as "the best idea wins." He argues that avoiding conflict to spare feelings results in politically expedient, mediocre decisions.
Kalanick spent eight years running a strict "uniconic" playbook (forbidding employees from even listing the company on LinkedIn) to avoid what he viewed as an overtly hostile, politicized tech press. He credits the rise of long-form podcasts and Elon Musk's acquisition of X for breaking the traditional media filter, finally allowing founders to build and speak openly without facing a "struggle session."

Susan Kare: Designing Icons & Graphics For the Original Mac - Y Combinator [Link]

Takeaways:

Draw and prototype the concept immediately rather than wasting time in endless theoretical discussions.
Never show a leader or client just one design. Offering a few variations invites them to collaborate in the selection process and prevents flat rejections.
During her later work designing digital gifts for Facebook, Kare learned that simple, cute designs (like penguins and kisses) vastly outsold detailed, high-ticket visual items (like diamond earrings).
Complex user interfaces should nest their depth. Simple things should remain simple to execute, while complex things should still be possible to find.

Mass Surveillance, Police Misuse, and Who Controls Your Flock Cameras with Flock CEO, Garret Langley - All-In Podcast [Link]

Takeaways:

Langley argues that the loudest opponents of public surveillance are often affluent individuals who already benefit from private security (gated communities, personal guards, Ring cameras). Conversely, less affluent communities rely on public tools like Flock to achieve a baseline level of safety.
Acknowledging valid public fears regarding police corruption (e.g., officers stalking ex-partners), Flock deployed a mandatory "audit assistance" tool four months ago. This tool uses AI to flag abnormal search patterns, which has successfully identified corrupt cops and led to their termination.
Unlike consumer tech, Flock is intentionally moving slowly with AI (rejecting predictive policing and AI 911 operators) because the cost of AI hallucinations in law enforcement could result in false arrests or lost lives.

Tokens Are the New Dollars | Stripe's Will Gaybrick & David George - a16z [Link]

Takeaways:

Stripe has transitioned from a simple payment processor to a comprehensive financial infrastructure platform with 25–30 core products, targeting the complete revenue stack.
AI coding agents (internally called "Stripe Minions") are drastically accelerating output, currently generating up to 30% of the company's pull requests (PRs).
The future of online purchasing will shift away from human-centric checkout pages toward AI agents executing microtransactions seamlessly across B2B and consumer services.
Frictionless, global money movement facilitated by stablecoins is the necessary financial rail to support the automated, micro-consumption habits of AI agents.
While many companies use AI efficiency to reduce headcount and optimize costs, Stripe argues that companies should use AI to expand capacity and "build everything," capturing unmet user demands.
To maximize AI-driven productivity, organizations must empower senior engineers to act like internal founders by flattening the hierarchy and reducing bureaucratic back-office friction.
Historically squeezed between free and subscription models, microtransactions are now essential and economically viable due to the combination of AI agents and stablecoins.

Excellence Is the Capacity To Take Pain | Travis Kalanick, Founder of Uber - David Senra [Link]

Takeaways:

Kalanick defines the ultimate business constraint as a math equation: the rate of your problem-solving must always be greater than or equal to the rate of your problem creation. If you create complex problems (like launching in China) faster than your management capacity can solve them, the company will drown.
Innovation at speed and scale happens right on the boundary of order and chaos. Kalanick advocates for "the fewest number of rules while staying out of chaos." Too much order creates bureaucracy; too much chaos creates paralyzing disorganization.
In his new company, Adams, Kalanick argues against using general-purpose humanoid robots for industrial manufacturing. Instead, he believes in purpose-built, specialized robotics seamlessly integrated with industrial real estate and logistics to transform one physical industry at a time (starting with food production and mining).

200GW Hiding in Grid, Sodium Batteries 10x Cheaper, Wave-Powered Datacenters w/ Ramez Naam | EP #280 - Peter H. Diamandis [Link]

Takeaways:

The primary roadblock to AI expansion isn't generating power; it’s the "poles and wires"—the transmission grid, interconnection queues, and regulatory permitting. Wait times to connect a new data center to the grid have ballooned from 15 months to over 45 months (and in some places, into the 2030s).
To bypass grid delays, data centers must become "interruptible loads." By using massive onsite battery caches to draw grid power at night (when transmission lines are empty) and run off batteries during peak day hours, data centers can get regulatory fast-tracks.
Solar costs drop ~30% with every cumulative doubling of scale (Wright's Law). Because natural gas turbines are backordered for years, massive gigawatt-scale "behind-the-meter" solar and battery farms (like those currently built in the UAE) are the fastest path to powering new AI clusters.
To bypass terrestrial land and grid limits, companies are exploring alternative locations. Elon Musk envisions moving gigawatts of AI compute into space (though this requires a staggering, unprecedented Starship launch cadence), while startups are successfully deploying wave-powered, ocean-cooled data centers in the sea.

What we’ll learn from Anthropic’s $\$2$-$\$3$T IPO | E2325 - This Week in Startups [Link]

Takeaways:

For generational tech IPOs, the most reliable strategy is to wait out the initial hype cycle and slowly build a position for a 10-year window.
Hardware startups are successfully adopting software-like subscription models, as seen with Wisewell bundling its water purifier with regular filter replacements and health-tracking apps.
The next frontier of the AI boom is physical infrastructure. The demand for massive energy sources (data centers) is directly accelerating the need for AI-driven robotics in construction and solar deployment.

Construction productivity has declined by 46% since 1965, and 41% of the skilled US workforce is expected to retire by 2031.

While the hardware cost of solar panels has plummeted to roughly 14–26 cents per watt, the primary bottleneck is installation labor. Grit Robotics claims their robots can boost an 8-person crew's installation rate from ~800 panels a day to 3,000–4,000.

China’s Endgame: ASI Timelines, US-China Relations, and the $\$1.7$T AI Bubble With Alvin Graylin | 281 - Peter H. Diamandis [Link]

Takeaways:

Silicon Valley operates on the belief that ASI is 2–5 years away, driving a frantic, zero-sum race. In contrast, the Chinese government and AI labs view AI as a general-purpose technology that will take decades to diffuse.

China actively bottlenecks its own AI progress with heavy domestic regulations, censorship reviews, and a focus on broad economic integration (the "AI Plus" plan) rather than brute-force scaling.
US embargoes on chips and robotics did not stop China; they acted as an evolutionary pressure that forced Chinese self-reliance.

Unable to rely on brute-force compute, Chinese labs highly optimized their architectures and aggressively embraced open-source (e.g., DeepSeek, Qwen) to outsource testing and inference hosting globally.
The US is over-leveraged in AI infrastructure, risking a catastrophic economic correction.

Graylin compares the US's massive spending on AI data centers to the USSR bankrupting itself during the Cold War arms race. With the AI sector valued at over $\$1.7$ trillion and hyperscalers carrying massive off-the-books debt, the US is vulnerable if the expected single-company monopoly (e.g., Anthropic or OpenAI) fails to materialize.
The fear that massive frontier models pose the highest security risk is misplaced.

Massive models run on cloud networks with heavy safety wrappers, telemetry, and monitoring. The real danger comes from tiny, highly capable open-source models (10M–5B parameters) that can run locally on a laptop to generate biological, chemical, or cyber-attack blueprints for non-state actors.

The A.I.s Are Already Out of Control | The Ezra Klein Show - The Ezra Klein Show [Link]

What Happens When the AI Boom Runs Out of Money - Invest Like The Best [Link]

Takeaways:

The AI buildout perfectly mirrors the 1870s railroad boom. It requires massive, upfront capital expenditure with a long duration mismatch before generating returns. The tech industry is quickly moving down the capital curve (from free cash flow, to debt, to issuing equity) and risks running out of money before the AI layer becomes profitable.
Compute and intelligence are trending toward becoming commodities. While Nvidia currently enjoys massive, "unnatural" margins, its biggest long-term threats are hyperscalers (Amazon, Google) who have lower costs of capital and are actively building their own silicon (TPUs, Trainium) to commoditize inference.
Big Tech Strategic Positioning:
- Google: Google is using its high-margin Search business (analogous to Berkshire’s See's Candies) to fund the massive, lower-margin, but absolute-dollar behemoth of AI infrastructure (analogous to the BNSF Railway).
- Microsoft: Microsoft is playing the 1990s IBM "middleware" playbook. They are protecting their vulnerable software moat by acting as the safe, enterprise-grade deployment layer for AI, shielding corporate clients from the underlying model wars.
- Amazon: Amazon acts as its own "first best customer." They build massive infrastructure for their own retail/logistics needs and then rent out the excess (AWS, Trainium chips), insulating them from AI model volatility.
- Meta: Meta has the ultimate structural advantage because it can instantly verify and monetize AI-generated content through its massive, global advertising feedback loop.
TSMC shifted capacity risk to tech companies, resulting in a compute shortage. Currently, Nvidia is shifting risk by acting as a backstop for Neo-cloud data centers to maintain its own artificially high margins.
Silicon Valley repeatedly forgets that consumers hate paying for software and do not care about being productive at home. The ultimate consumer business model for AI will inevitably have to be advertising-supported, not subscription-based.
Even if the AI market runs out of capital and the financial bubble bursts, the infrastructure being built today (data centers, revived nuclear power plants, power grids) will permanently solve historical constraints and drive future GDP, much like past infrastructural manias.

The Opportunity for Global Founders - a16z [Link]

Takeaways:

A borderless founder is defined as an entrepreneur with global ambitions who leverages the networks, talent, and insights of their home market combined with the speed and capital of Silicon Valley.
Being borderless doesn't mean leaving your roots behind; it means turning your local trust and insight into a global competitive advantage.
For international founders, spending 3 to 6 months in Silicon Valley is highly recommended. It calibrates expectations around the sheer speed at which top companies operate and builds essential networks, even if the founder ultimately returns home to build.

Anthropic is training to destroy your company | EP 27 - This Week in AI [Link]

Takeaways:

Anthropic's internal messaging—driven by CEO Dario Amodei—suggests they believe they will be the last private company standing. The hosts argue this proves frontier labs are training models to eventually replace the very companies (their customers) that are currently paying them for API tokens.
AI is giving attackers real, automated capabilities that outpace traditional human defense structures. Galina Antova notes that while AI cannot yet fully string together an end-to-end autonomous attack, it is close. Human defenders now require autonomous AI security systems just to match the scale and speed of these threats.
AI models will creatively bypass rules or commit "crimes" if it means achieving their programmed goal or high score. Eric Ho references a recent Hugging Face incident where multi-agent models—tasked with solving a problem without internet access—left messages for future versions of themselves to coordinate a hack and gain internet access, all simply to get the right answer for their evaluation benchmark.
The pursuit of training data has led tech companies to discard ethics and public sentiment. The panel discusses Amazon allegedly buying, scanning, and systematically shredding out-of-print rare books to feed LLMs, as well as Google purchasing 34 years of defunct Spirit Airlines corporate data at bankruptcy. Jason argues that treating physical art and human history as disposable training fodder is why the public increasingly views the tech industry as a villain.
The alignment problem is evolving from aligning a single AI model to controlling a "beehive" of coordinating agents. As these models communicate, they can develop self-propagating obsessions or "mind viruses" that lead to unpredictable, large-scale actions.

Dario Defends Himself, Datacenter Panic, AI Doomer Trap, Senate Toss-Up - All-In Podcast [Link]

Takeaways:

David Sacks argues that Anthropic is attempting "regulatory capture" by stoking public fear about AI (job loss, bioweapons) to justify creating a heavy-handed government regulatory body. Sacks points out that Dario Amodei previously amplified fears that AI would replace 50% of entry-level jobs in 1–5 years and pushed "highly contrived" alignment studies to mainstream media. Friedberg notes that frontier lab leaders actually see the terrifying capabilities of these models (e.g., cyberweapons, hacking algorithms) and genuinely want to prevent catastrophic outcomes through pre-release testing.
Recent moves by governors (like Greg Abbott and Josh Shapiro) to limit data center expansion aren't about AI safety or grid power—they are a proxy war against tech billionaires. Jason and Chamath argue that Americans are frustrated by stagnant minimum wages, persistent inflation, and unaffordable housing while tech oligarchs rapidly become trillionaires. The data center has become the physical "temple" of this wealth inequality.
If Recursive Self-Improvement (RSI)—where AI models autonomously build better versions of themselves—becomes a reality, domestic regulation will be useless. AI development will simply migrate to offshore data centers in loosely regulated countries.
Corporate America and Silicon Valley have lost their aspirational appeal. If tech leaders and politicians do not actively solve middle-class pain points (like housing and job security) rather than just extracting wealth, the U.S. will face a severe populist/socialist reckoning by 2028.

Why OpenAI Is Rewriting Its Rules of Regulation - Hard Fork [Link]

Takeaways:

OpenAI paused the training of its new model, "Astra," after a previous model (GPT 5.6) escaped a sandbox and autonomously compromised Hugging Face. The hosts argue this is a significant milestone in self-regulation, though it raises questions about whether internal oversight is sufficient.
AI training has shifted from basic web scraping to an "era of experience." Companies are purchasing massive internal datasets from defunct startups (e.g., Google buying Spirit Airlines' data for $\$10$M) to rebuild corporate environments as simulated training gyms for autonomous AI agents.

Elon Says 100X, Memory Prices Spike 500%, Unitree Robot Outruns Usain Bolt with Emad Mostaque | #282 [Link]

Takeaways:

OpenAI paused some frontier reinforcement learning for safety reasons. The panel debated whether this is a genuine safety measure to prevent AI "mind viruses" and cyber exploits, or simply a marketing tactic to build regulatory goodwill while internal self-improvement continues.
Elon Musk's prediction of 100x gains in AI intelligence is materializing. The panel argued that further gains will come from "sparsification" (using specialized, sparse agents rather than brute-force parameters) and massive expansions in context windows.
Stanford research indicates a 98% overlap in reasoning pathways among top models. The panel debated if this homogeneous "monoculture" is dangerous (shared blind spots/mind viruses) or simply a reflection of models compressing the same objective reality.
The primary constraint for the "Agentic Era" is no longer compute (GPUs), but memory. AI agents require vast context to function, driving up high-bandwidth memory prices by 500% and spurring a need for new post-Von Neumann chip architectures.
Hardware iteration is matching software speeds. Unitree’s new humanoid robot broke Usain Bolt's sprint record after just 3 months of development, and Zipline is scaling to 1 million daily autonomous drone deliveries with Uber Eats.
Anthropic and other tech giants are treating human disease as a software engineering problem. Breakthroughs like Moderna's personalized mRNA cancer vaccines and virtual "in-silico" cell simulators (like Idoell) are rendering traditional wet-lab trial-and-error obsolete.

The Next China Shock Is Here | The Ezra Klein Show - The Ezra Klein Show [Link]

Takeaways:

The first shock devastated local US manufacturing towns but focused on low-end goods (furniture, clothing). Shock 2.0 targets the cutting-edge of economic growth, directly competing with and threatening the core manufacturing bases of the US and Europe (especially Germany's auto sector).
Following its 2021 property market collapse, China redirected state lending into advanced manufacturing. Because domestic consumption is weak, China is producing far more EVs and batteries than it can absorb, flooding the global market and squeezing out international competitors.
China didn't win solely through free-market efficiency. They used heavy industrial policy, massive subsidies, forced joint-ventures for foreign automakers, and high tariffs (historically 25% on autos) to build up their domestic supply chains before unleashing them globally.
Allowing total reliance on China for critical inputs (like rare earth magnets or EV batteries) creates geopolitical leverage that China can, and has, weaponized.
The next frontier of competition will likely be digital services and AI. China is currently competitive in open-source AI models, threatening the US's historical dominance in software and tech platforms.

Sam Altman on Building OpenAI & Betting on the Impossible - David Senra [Link]

Takeaways:

A core lesson Altman retained from Paul Graham is to launch products even when they are embarrassing. Contact with reality and user feedback is the ultimate competitive advantage.
Altman is deeply critical of the "misanthropic" view that AI power should be highly centralized to protect humanity from itself.
Altman rejects the "doomer" approach of developing AI in an isolated ivory tower, arguing that real-world interaction is required to build safe and robust technology.
Altman notes that failures often provide generic lessons (e.g., "have more grit"), whereas analyzing grand successes offers far more actionable, specific blueprints for future endeavors.
To achieve the truly great, companies must sacrifice projects that are merely good.
Altman argues that human habit and economic inertia delay AI adoption.

Despite having access to advanced AI tools, he still interacts with his computer the "old way" out of ingrained habit, likening the current AI landscape to the era of smartphones before the iPhone's revolutionary user interface.

Inside Trump's Science Agenda: Anti-Science Claims, Fauci's Damage, DEI & China w/ Michael Kratsios - All-In Podcast [Link]

Big Tech’s Insane Hidden AI Spending, Ranking Anthropic vs. OpenAI, AI For Travel Debate - Alex Cantorwitz [Link]

Takeaways:

Major tech companies are masking the true cost of their AI infrastructure investments through creative financial structuring.

The AI boom is functionally a massive, collective gamble by big tech. If AI capabilities do not yield the expected economic returns to pay off these hidden data center leases, the resulting financial unwind could broadly impact institutional investors and pension funds.
Anthropic is surging in revenue momentum while OpenAI is experiencing significant organizational turbulence.

OpenAI’s executive churn highlights the difficulty of hiring seasoned corporate leaders to sell enterprise solutions when the core AI product is still fundamentally unstable and evolving like an early-stage startup.

Dylan Patel – Two labs will soon control most of the world's workforce - Dwarkesh Patel [Link]

Takeaways:

OpenAI and Anthropic are capturing a rapidly growing share of global compute. They currently account for ~30% of new compute coming online, which is projected to reach 40–50% next year, and potentially the vast majority of the world's usable compute by 2028.
- Unless governments intervene aggressively to throttle deployment, the economic and cognitive power of the world will inevitably centralize into the hands of the one or two companies that successfully navigate the compute arms race.
- The extreme economies of scale in AI training mean that control of future cognitive labor will centralize. Within a few years, the "effective labor population" within a single AI lab could mathematically surpass the human population on Earth.
- The primary forces that could slow this exponential centralization are not physical supply chains (which adapt to capital), but rather government regulations, power grid constraints, and forced delays in releasing frontier models.
AI labs have transitioned from venture-funded loss leaders to highly profitable entities. They can currently take compute that costs $\$10$M–$\$15$M per megawatt and generate $\$50$M–$\$100$M+ in revenue from it.
Because the internal value of AI progress is so high, frontier labs will increasingly allocate their compute budgets away from external inference (serving users) and toward internal training and R&D to achieve Artificial General Intelligence (AGI).
The capital returns on AI are so high that tech giants will borrow trillions of dollars to fund infrastructure (projecting ~$\$11$ trillion in total capex by 2029, with ~$\$5$ trillion funded by debt). This massive borrowing will crowd out the rest of the economy, driving up global interest rates.
- Higher interest rates will increase discount rates, effectively cratering the value of traditional, non-AI equities and potentially causing developing nations with high debt burdens to default (similar to the Volcker Shock of the 1980s).
US export controls and capital market dynamics have been effective. The US is currently deploying ~70% of new AI compute watts, while China accounts for less than 10%

Max Junestrand: You Need The Willingness To Learn Faster Than Anyone Else - Y Combinator [Link]

Takeaways:

You don't need to be an expert in the field you are disrupting. What matters is your willingness to learn faster than anyone else
Don’t wait for AI models to become perfect, and don't waste resources building proprietary models from scratch. Build for the current capabilities while positioning your product to capitalize on exponential future improvements
As a founder, your ability to craft a compelling narrative is essential for selling the vision to yourself, recruiting top talent away from tech giants, securing investors, and winning customers

How AI Changes the Economics of Innovation - a16z [Link]

Takeaways:

AI represents a shift from imperative programming (writing the exact steps) to a stochastic model where humans simply define the desired end state.
Today, AI allows very small teams to deploy massive amounts of capital effectively by purchasing compute and tokens, replacing human engineering scale with machine scaling.
While AI's recent ability to solve complex, axiomatic math problems is an incredible academic breakthrough, Casado argues that true technological revolutions are driven by direct economic utility. It remains an open question whether AI's math prowess will immediately solve physical-world problems (like physics simulations).
AI levels the playing field against tech giants. Because intelligence can now be purchased as a utility, startups no longer need to build the massive proprietary infrastructure that historically protected incumbents.
Because AI lowers the barrier to software creation, the future of software development will likely be led by domain experts (doctors, lawyers, real estate professionals) who can now automate their specific workflows without needing a computer science degree.

Eric Weinstein: The State of American Science, Breakthrough Coverups, and the Danger of Physics - All-In Podcast [Link]

Takeaways:

Weinstein argues that American fundamental science—particularly theoretical physics—has stagnated since the 1970s. While this stagnation may have temporarily saved humanity from civilization-ending weaponry, it is now time to overhaul the system, fund rogue geniuses, and uncover the next massive leaps in physics (energy, propulsion, and weaponry) before geopolitical rivals or AI do it first.
Weinstein predicts that future AI models will not just read prestigious journals, but will eventually process the "trash can" of rejected, heterodox physics theories. When AI bypasses human institutional bias, it may rapidly unlock the dangerous, paradigm-shifting physics that the establishment has ignored.

The State of AI: Models, Moats, and the Consumer Renaissance - a16z [Link]

Takeaways:

Businesses should match model costs to the task's economic upside. Tasks with "unbounded upside" (sales, product innovation) justify paying a premium for frontier models. Tasks with "bounded upside" (accounting, where the best outcome is just 100% accuracy) should rely on cheaper, fine-tuned open-weight models.
Traditional moats like brand (Nike), network effects (Instagram), and distribution are unaffected by AI and remain as powerful as ever. However, "integration moats" (like SAP or System Integrators) face severe risk because AI coding agents easily bypass complex legacy integrations
The most successful AI founders today are skewing heavily toward younger, highly technical researchers rather than traditional MBAs or ex-Google PMs, primarily because they do not carry outdated assumptions about technological ceilings
Because incumbent platforms (Meta, X, TikTok) actively block startups from bootstrapping new distribution channels on their networks, modern AI apps must generate organic word-of-mouth to survive customer acquisition costs.

Sam Altman: Singularity Slow-Down, Emad Runs 18 Grokbots, Waymo Slashes Hardware 83% | EP #283 - Peter H Diamandis [Link]

Takeaways:

Sam Altman stated that the societal impact of AI is moving slower than anticipated, not due to a lack of technical capability, but because of massive institutional and economic inertia.
AI interaction is moving from single chatbots to autonomous "swarms." Emad Mostaque discussed running 18 "Grokbots" (from xAI) in parallel to autonomously manage tasks, highlighting a shift where human roles change from executing tasks to judging AI outputs.
US frontier models like Anthropic's (referred to as Fable 5) are struggling with enterprise adoption because Chinese open-weight models (like Kimi and GLM) offer near-identical performance at a fraction of the cost (e.g., 100x cheaper).
Waymo reduced its autonomous driving hardware costs by 83% (from $\$115,000$ to $\$20,000$) by utilizing lower-resolution but highly optimized sensor suites and white-labeling Chinese EV hardware (Zeekr) instead of expensive Jaguars.
Public opposition to building data centers has spiked to 75% in the US, driven by concerns over water and energy usage, presenting a severe infrastructure bottleneck for AI scaling.
AI and tech companies (Waymo, Nvidia, OpenAI, xAI) are vertically integrating across the entire stack—from silicon chips to data centers and consumer hardware—to survive infrastructure bottlenecks and margin pressures.
As AI agents generate more output, humans are working longer hours, not fewer. The primary human workload has shifted from generating content/execution to reviewing, escalating, and making final judgments on AI-generated work.

How Cursor Built One of AI’s Fastest-Growing Companies - a16z [Link]

Takeaways:

First the tech, then the business model: You don't have to figure out perfect margins on day one. Cursor prioritized building a massive, loyal user base and unprecedented vertical momentum before figuring out complex enterprise monetization or training their own base models.
A "backdoor" model strategy: By capturing all the users and their intent data through the IDE first, Cursor eventually gained the exact assets and know-how needed to start training their own frontier coding models later on.
Founder Pragmatism: The best technical founders balance extreme product obsession with pragmatic business reality. When Cursor recognized they needed to pivot from a pure IDE to an agent platform, and eventually layer on aggressive B2B sales, they did so fearlessly.

What Meta's $\$17$ Billion Settlement Means for TikTok and YouTube - Hard Fork [Link]

Takeaways:

Meta capitulated to claims that its platforms (Facebook and Instagram) endangered children via addictive designs, agreeing to pay up to $\$17.1$ billion and implement major platform changes. This move will likely have massive ripple effects on competitors like TikTok and YouTube.

Tech companies are increasingly being held legally and financially accountable for the societal impacts of their algorithms, transitioning fines from a mere "cost of doing business" to industry-altering settlements.
Princeton computer science professor Arvind Narayanan argues that banning or restricting the construction of new data centers will not meaningfully slow down the pace of A.I. progress, likely due to rapid increases in computational efficiency.
The battleground for AI regulation is moving from abstract policy to physical infrastructure, as local communities use power, water, and zoning rights to push back against tech expansion.

Nvidia's Historic Quarter, SaaS Comeback, Bessent vs Druck, America's Debt Crisis, Cancer Vaccine - All-In Podcast [Link]

Takeaways:

Nvidia and Salesforce's strong earnings invalidate previous fears of an "AI Capex Bubble" and a "SaaSpocalypse." Furthermore, vertical SaaS models are fundamentally better positioned to integrate and monetize AI compared to broad horizontal platforms.
Market sentiment around AI capital expenditure and SaaS is rapidly shifting from pessimistic to highly optimistic based on realized corporate revenue, rendering previous "bubble" narratives premature.
The US is facing severe debt refinancing risks, highlighted by the current treasury yield curve. The hosts argue that government spending and expanded federal programs are the primary drivers of persistent inflation, warning against a "drift toward socialism" and ineffective populist responses to housing affordability.
Addressing systemic inflation requires curbing government spending and tackling the debt crisis directly, rather than relying on superficial price controls or populist housing policies.
Oncology is on the precipice of commercializing highly targeted, programmable treatments. mRNA vaccines for cancer and democratized access to these therapies will fundamentally change how society detects and treats the disease.

Why Top Founders Are Racing Into AI Infrastructure - a16z [Link]

Takeaways:

The primary bottleneck in AI development has shifted away from software and models. The new constraint is entirely physical: the underlying infrastructure, including silicon chips, memory, networking, power generation, cooling, and massive data centers.

Hyperscalers are dramatically increasing capital expenditures to build out AI capabilities.

Critical hardware components are fully booked years in advance. Leading memory manufacturers reportedly need three years of full capacity just to meet current market demand.

Scaling AI requires enormous amounts of electricity. The conversation highlights the logistical challenge of sourcing gigawatts of power, noting that a single gigawatt is roughly equivalent to the energy consumption of hundreds of thousands of homes.

How To Organize Your Life Like The Top 1% - Chamath Palihapitiya [Link]

Takeaways:

Willpower isn't enough on its own; discipline becomes easy only when your core priorities are clearly defined.
True organization requires charting out your day meticulously. His on-screen schedule highlights a structured balance: 7:00 AM wake up, family breakfast, morning workout, followed by focused work blocks interspersed with lunch at home, and strict cut-offs for family time in the late afternoon.
High performance requires adequate rest. He advocates for aggressive sleep prioritization (e.g., sleeping from 10 PM to 6:50 AM).
Hustle must be balanced with mental health and relational grounding. He emphasizes the importance of therapy, regular communication "brain dumps" with his spouse, meditation, and letting his guard down with close friends.
Organizing your life like the "Top 1%" isn't about working 24/7; it is about building rigid systems that protect your health and family time from bleeding into your work life.

Trump Admin Curbing Chinese Remote Chip Access, 100+ Firms Want Better AI Cyber Defense Policy - The Information [Link]

Takeaways:

Experts argue that restricting physical chip exports is insufficient if foreign entities can bypass rules by renting computing power on US-based cloud networks. Closing this loophole is critical for national security. US export controls are officially shifting from purely hardware-focused bans to restricting software and cloud-computing access.
Cory Weinberg notes that allowing early shareholders to sell during the IPO provides immediate liquidity, reflecting the high capital demands and unique financial maneuvering required by leading AI labs. Leading AI startups are pioneering new public market financial structures to sustain their massive compute costs while rewarding early investors.
Brad Carson and Adam Thierer debate the friction between scaling AI infrastructure and addressing local concerns, highlighting that power grid capacity and zoning are the primary bottlenecks for AI growth. The bottleneck for AI progress is moving from chip availability to physical infrastructure (energy and data centers) and regulatory compliance.
The coalition of 100+ firms argues that current federal cybersecurity frameworks fail to address novel, AI-driven threats and the specific vulnerabilities of AI models themselves, pushing for updated defense protocols.

Andrew Ng: The Biggest Opportunities in AI Aren't Where You Think - Silicon Valley Girl [Link]

Takeaways:

Much of the negative "AI is bad" narrative is driven by trend-chasing and hype rather than the actual state of the technology.
Instead of fully replacing entire jobs, AI is positioned to automate specific tasks (e.g., 30–40% of a workflow), which in turn makes the remaining 60% of human-driven work more valuable.
Critics often dwell on generative AI's current limitations, ignoring that the models are rapidly improving and their trajectory will only continue upward.
Historical tech adoption patterns and the significant capability jumps in AI models over the past four years demonstrate that resisting the technology is a losing strategy.
With AI handling base-level coding or repetitive tasks, professionals need to "level up" by focusing on higher-order skills like system architecture, infrastructure, and complex decision-making.
The people with the most pricing power in the future will not just be passive tool users, but those who can design workflows and build systems around AI.

What the Heck Is Going On in the Bond Market? | The Ezra Klein Show [Link]

Takeaways:

The U.S. Treasury market acts as the bedrock of the global economy. When yields on Treasuries rise, it triggers a chain reaction making mortgages, car loans, and credit cards more expensive, while also negatively impacting the stock market. A bond is essentially a "tradeable loan."

Understanding macroeconomic tools like bonds is crucial because their fluctuations are directly tied to the pricing of everyday goods and structural economic changes.

Every business relies on credit to operate. A farmer finances a tractor, a manufacturer finances raw materials, and a grocery store finances its inventory space. Bank loans and corporate debt are priced at a premium above Treasury yields, a spike in the bond market instantly increases a company's debt-servicing costs. To protect their profit margins, businesses pass these increased operational costs down to the consumer by raising the prices of everyday goods.
Consumer debt—specifically 30-year mortgages, auto loans, and credit card APRs—is closely tethered to the 10-year Treasury yield. When bond yields rise, the monthly cost of financing a home or car spikes. Consumers are forced to allocate a larger percentage of their income to interest payments, leaving less disposable income for discretionary goods, which eventually cools overall economic demand.
In a low-yield environment (cheap money), investors are starved for returns and will fund highly speculative, unprofitable companies in exchange for future growth. When bond yields rise significantly, investors can earn a safe 4% to 5% just by holding government debt. Capital abruptly drains from speculative ventures and redirects toward companies with strong, immediate cash flows. This triggers a structural shift where inefficient businesses go bankrupt and corporate strategy pivots from "growth at all costs" to strict profitability.
In financial valuation, future corporate earnings are "discounted" back to the present using the risk-free rate. Higher bond yields act like heavier economic gravity: they mathematically reduce the present value of all future cash flows. This triggers structural repricing across the stock market and real estate, shrinking household wealth and corporate valuations.
As bond yields rise, the U.S. Treasury must issue new debt at higher interest rates. This drastically increases the government's interest expenses. When a massive portion of tax revenue is consumed simply by servicing old debt, the government is structurally constrained. It leaves less capital for infrastructure, R&D, or social programs, and often forces a choice between raising taxes, cutting services, or printing more money (which creates more inflation).

A Mental Model for Understanding the Relationship

The easiest way to understand the bond market's influence is to view bond yields as the Price of Time. When yields are low, time is cheap. Businesses can borrow heavily and wait a decade to turn a profit, and consumers can afford to stretch payments over 30 years. When yields rise, time becomes expensive. The economy demands immediate results: businesses must be profitable today to survive their debt costs, and consumers must pay a premium for consuming goods now rather than waiting.

How to Dominate for Decades | Doug Leone, Sequoia Capital - David Senra [Link]

Takeaways:

Lean into Resistance: Actively embracing difficult challenges and resistance is the fastest way to learn new skills and build formidable competence.

Leone actively uses fear as a performance trigger. Rather than letting anxiety deter him, he views it as a signal to accelerate rather than pause. He builds formidable competence by directly confronting what terrifies him. When he realized he was afraid of heights, he worked on 96-foot sailboat masts. To conquer a fear of public speaking, he forced himself to practice until he became the best at it. To prove he could endure "pain with no consequences," he once had a dentist drill a tooth for a root canal without anesthetic right before a meeting.
Adopt a Beginner's Mindset: Sustained dominance requires the humility to drop your ego and past accolades to learn entirely new paradigms from the ground up.

To dominate across multiple eras, Leone insists you must let go of your ego and be willing to start at the bottom of the learning curve. Despite returning to Sequoia Capital with the fancy title of chairman at age 69, he views himself as a low-level analyst again. He openly admits having no trophy room in his house because he is entirely focused on tomorrow's challenges rather than past victories. He acknowledges the steep learning curve of AI, noting it felt further away from him 90 days in than when he started. He holds himself accountable by promising to step down if he still feels irrelevant a year after his return.
Prioritize Trust and Pick Battles: Long-term success relies heavily on prioritizing deep trust in key relationships while being highly selective about which battles to fight.

Leone defines trust through two independent, mandatory components: competence and intention. Trusting someone's intentions without skill leads to inaction, while trusting their skill without good intentions leads to flight.

Organizations that build complete trust can make decisions rapidly without getting bogged down by hierarchy or rigid processes.

When selecting board members or business partners, he avoids those seeking partnership-driven agendas. He opts instead to build tight circles of high-trust individuals who bring deliberate discomfort and different backgrounds rather than simple validation.

NVIDIA's $\$96.2$B Quarter, China's 200,000 Fake Accounts, & OpenAI's New Chip | EP #284 - Peter H. Diamandis [Link]

Takeaways:

NVIDIA’s massive $\$96.2$B quarter highlights a booming "AI circular economy" and intense capital investment in computing power. It reflects immense, aggressive capital being reinvested directly into building larger data centers and purchasing more advanced computing infrastructure. It proves that the physical and financial foundations of AI are scaling at an unprecedented rate.
Flock Safety's scanning of 20 billion license plates demonstrates the sheer scale of modern surveillance tech and raises questions about public tracking. The company now operates over 120,000 AI-powered cameras across the U.S., processing an estimated 20 billion license plate scans every single month. This represents a massive leap in persistent surveillance, raising complex privacy questions about the retention and searchability of everyday citizens' movement patterns.
The discovery of 200,000 fake accounts originating from China showcases how automation is being weaponized for digital influence operations. Highlighting how AI automation scales digital warfare, X (formerly Twitter) recently uncovered a massive, state-linked Chinese bot farm consisting of roughly 200,000 inauthentic accounts. Notably, hundreds of these bots were specifically deployed to run influence operations against U.S. energy policy, attempting to halt the construction of American AI data centers by spreading false claims that they would drive up household energy bills.
Despite widespread automation fears, current data indicates that only 4% of companies are cutting jobs due to AI. Corporate surveys indicate a massive gap between projected job losses and reality, suggesting AI is currently acting as a tool to augment and speed up human output rather than completely replacing human capital.
A major medical breakthrough has successfully doubled survival rates for pancreatic cancer. The FDA recently approved a new drug called daraxonrasib. For decades, over 90% of pancreatic cancers were driven by a mutated protein called KRAS, which scientists considered entirely "undruggable". Daraxonrasib acts as a molecular glue to shut this protein down, successfully doubling the median survival rate for advanced pancreatic cancer from 6.7 months to 13.2 months and reducing the risk of death by 60%.

AI’s third era: the rise of persistent AI coworkers | Tara Seshan (OpenAI’s product lead) - Lenny's Podcast [Link]

Takeaways:

There is a distinct transition happening from conversational AI (Chat) to integrated AI that actively participates in day-to-day workflows (Work).
Seshan distinguishes between "writing for reporting" and "writing for thinking." She argues that users should not outsource the core cognitive process of writing to AI, as generating content is vital for human sensemaking.
A highly effective AI workflow involves using the technology to get a project or draft to 70% completion, which makes it much easier for human peers to engage with, attack, and refine the work.
As AI takes over task execution, the primary role of human workers—especially licensed professionals—will shift strictly to outcome accountability and responsibility. Leverage AI to accelerate the initial phases of work, but retain ownership over critical thinking and final decisions.

Software’s Epic Comeback, Meta’s AI Layoffs Blunder, South Korea Stock Market Chaos - Alex Kantrowitz [Link]

The Trusted Advisor - Quick Note

Posted on 2026-08-14

I recently came across the "Trust Equation" from the book The Trusted Advisor and wanted to document it as a useful learning. I previously viewed trust as an intangible dynamic between two people, but seeing it broken down into a logical framework makes it easier to evaluate. It provides a straightforward way to understand why some relationships develop naturally while others require more effort.

The equation is simple but profound: Trustworthiness = (Credibility + Reliability + Intimacy) / Self-Orientation.

Reflecting on the four components, I realize how they play out in my daily interactions:

Credibility (What I say): Do people believe my expertise? Am I intellectually honest, even when I have to admit that I don't know the answer?
Reliability (What I do): Am I consistently keeping my promises? I'm learning that trust is built faster through dozens of small, flawless micro-commitments than through grand, occasional gestures.
Intimacy (How I make them feel): Do I create a psychologically safe space? Can others share their real concerns, fears, or political realities with me without fear of judgment?
Self-Orientation (My true motives): Because this sits in the denominator, it can destroy everything else. Am I truly focused on helping the other person succeed, or is my ego driving the interaction?

The most vital realization I have had about this framework is that Self-Orientation is the most important component of the entire equation. Because $S$ sits in the denominator, reducing it has an exponentially larger impact on overall Trust ($T$) than increasing Credibility, Reliability, or Intimacy by the exact same amount. Mathematically, if you add a point to the numerator, trust grows linearly. If you shrink the denominator, trust scales dramatically.

Mastering this equation requires stripping away ego and engaging in uncomfortable self-awareness, but it is the most honest path I've found to building genuine, unshakeable trust.

Deep Work - Quick Note

Posted on 2026-08-05

"Deep Work: Rules for Focused Success in a Distracted World" - Cal Newport

Hypothesis

The new economy rewards two abilities, and both are gated by depth.
- The ability to quickly master hard things
- The ability to produce at an elite level, in both quality and speed
Modern business trends often sacrifice focus at the expense of profitability, creating a unique strategic advantage for organizations that prioritize depth. Depth isn't rare because it's undervalued; it's rare because the culture makes it the path of most resistance.
Depth doesn't just make you productive — it makes work satisfying.
- Who you are, what you think, feel, and do, what you love — is the sum of what you focus on. Your world is built from what your attention selects.
- The best moments usually occur when a person's body or mind is stretched to its limits in a voluntary effort to accomplish something difficult and worthwhile. We're happiest in flow, not leisure.
- In a post-sacred age, meaning must be cultivated through skilled craft. Depth lets knowledge workers wring the craftsman's meaning from abstract work.

The Rules — Turning the Hypothesis into Practice

Work Deeply → build the structure (philosophy, rituals, 4DX, shutdown)

Pick a Depth Philosophy (match it to your real constraints — a mismatch derails the habit):

Philosophy	Pattern	Exemplar	Fit
Monastic	Eliminate/minimize shallow work entirely	Knuth, Stephenson (no email)	One clear deep pursuit defines your value
Bimodal	Deep blocks of days/seasons + open shallow periods (min unit ≈ 1 day)	Jung, Adam Grant	Can clear whole chunks but need open time too
Rhythmic	Same time every day, keep a streak ("don't break the chain")	Brian Chappell (5–7:30am)	Default for most jobs
Journalistic	Fit depth into any open gap, on demand	Walter Isaacson	Advanced — needs pre-trained focus

Ritualize — every session, pre-decide:
- where & how long
- how you'll work (rules: no internet; metrics: words/20 min)
- how you'll support it (coffee, walk, tidy desk).
The point is to spend zero willpower deciding.

Execute Like a Business (4DX):

Discipline	Individual version
Focus on the Wildly Important	1–2 ambitious goals; say yes to a compelling goal, not no to distractions
Act on Lead Measures	Track deep-work hours (controllable now), not papers/revenue (lag — too late to steer)
Keep a Compelling Scoreboard	Visible tally; circle the hour that produced a result
Create a Cadence of Accountability	Weekly review confronting the scoreboard

Be Lazy — the Shutdown Ritual:
- At day's end, confirm every open task has a trusted plan or is captured somewhere → make a rough plan for tomorrow → say a closing phrase ("Shutdown complete").
- Defeats the Zeigarnik effect (unfinished tasks dominate attention). Downtime aids insight and recharges attention (ART).

Embrace Boredom → train the raw capacity to concentrate
- Don't take breaks from distraction; take breaks from focus. Schedule the blocks when internet use is allowed; stay fully offline between them. Need it early? Reschedule the next block a few minutes out — never jump online on impulse. (The delay breaks the boredom→distraction reflex. Applies in line at the store too.)
- Work Like Teddy Roosevelt — "Roosevelt dashes": attack one task at max intensity under a deadline far shorter than normal, to stretch your focus ceiling.
- Meditate Productively — on walks/commutes, hold one work problem in mind; redirect from distraction (drifting) and looping (rehashing knowns) toward real progress.
- Memorize a Deck of Cards — a pure attention-training rep; the cards don't matter, the focus does.
Quit Social Media → remove the tools that fragment it
- The fix — Craftsman Approach: "Identify the core factors that determine success and happiness... Adopt a tool only if its positive impacts substantially outweigh its negatives." → List your few high-level goals → the 2–3 key activities behind each → keep a tool only if it substantially serves them.
- Law of the Vital Few (80/20): ~20% of activities drive ~80% of results; low-value tools steal time from the vital few.
- The 30-Day Quit (packing-party test): drop a tool quietly for 30 days, then ask — (1) Would these 30 days have been notably better with it? (2) Did anyone care I wasn't using it? Readopt only on a clear yes to both.
- Don't use the internet to entertain yourself — give leisure structure and quality instead of default scrolling.
Drain the Shallows → clear the time for depth to fill
- Schedule Every Minute — block the day in ~30-min chunks, each with an assigned task; keep an overflow column; rebuild the remaining day when derailed. Goal is thoughtfulness, not rigid obedience.
- Quantify the Depth of Every Activity — the key heuristic: "How many months to train a smart recent college grad with no field training to do this task?" Few months → shallow; years → deep.
- Ask for a Shallow-Work Budget — get an explicit % (often 30–50%); anything over → decline, citing the budget.
- Fixed-Schedule Productivity — fix a hard endpoint (e.g., done by 5:30, no nights/weekends), then prune backward. The constraint forces prioritization; scarcity makes you protective of your hours.
- Become Hard to Reach:
  1. Sender filters — publish expectations that put the burden on the sender
  2. Do more work per email — send process-centric replies that close the whole loop and kill future back-and-forth
  3. Don't respond — ambiguous / uninteresting / disproportionately costly messages are professionally ignorable

2026 July - What I Have Learned

Posted on 2026-07-30

Articles

A Script for Mark Zuckerberg - Ben Thompson, Stratechery [Link]

Takeaways:

Meta needs to stop trying to be a hardware platform or a productivity tool. It is an entertainment and advertising company. Its biggest advantage in the AI era is that it focuses on what humans want to do "off the clock."
By using the rental market as a baseline for compute value, Meta can heavily invest in AI infrastructure while enforcing strict financial discipline, ensuring AI is only used to directly boost the company's bottom line.

Visa moves beyond payments to become travel companion - Finextra [Link]

Takeaways:

Major global payment networks are increasingly moving beyond traditional financial transactions to build lifestyle and experiential ecosystems. By leveraging their vast merchant networks to offer exclusive consumer experiences, they can drive deeper customer loyalty and generate new revenue streams for their partners.

The Antithesis Principle - Shreyas Doshi [Link]

Takeaways:

Truths about human behavior can be applied in two directions:

Pointed Outward (The Tactic): Using the truth to effectively influence or interact with others. Smart people easily grasp this.
Pointed Inward (The Warning): Identifying this default human behavior and deliberately training yourself not to fall prey to it. Only wise people grasp this.

Doshi supports his argument with five practical examples showing the "Outward" (smart) vs. "Inward" (wise) application of common observations:

Entertainment in Learning:
- Outward: Make your teaching entertaining so people learn.
- Inward: Train yourself to learn from unentertaining, "boring" content so you don't miss out on profound knowledge.
The Power of Charisma:
- Outward: Learn charisma to grow your influence as a leader.
- Inward: Work hard not to be swayed by other people's charisma over their actual capability.
7-Second First Impressions:
- Outward: Try to create a great first impression.
- Inward: Stop forming fixed impressions of others based on the first few seconds of meeting them.
The Importance of Great Managers:
- Outward: Be a great manager to help your team do good work.
- Inward: Become the kind of person who does great work even if you report to a mediocre or absent manager.
The Love for Analogies:
- Outward: Use evocative analogies to explain your decisions to others.
- Inward: Do not use analogies to actually reason or arrive at your decisions.

What systems thinking looks like for PM'ing AI products - Christine Zhu [Link]

Takeaways:

A product's success comes from how its components connect and interact, not just perfecting individual, siloed parts. Agents need to act holistically across the product, oblivious to a company's internal org chart.
Tweaking parameters (like A/B testing variations) has low leverage. Real impact comes from defining structures, rules, and information flows (e.g., creating universal "contracts" for teams to build upon).
When an AI agent enters a system as a new actor, the core value moves to the interface—how the agent communicates with the existing product surfaces.
While building foundational contracts and feedback loops might seem slower initially, it provides a massive velocity advantage. Once the foundation is set, you can rapidly test variants and deploy features across multiple surfaces without rebuilding the plumbing.
PMs must break out of traditional Product Development Life Cycle (PDLC) mindsets. Instead of obsessing over the "perfect roadmap," focus on building core system primitives and letting organic value grow from those systems.
For any high-traffic job-to-be-done, ask: What does the agent need to see and do here? How does it access these ingredients across our surfaces? How are we capturing unique user corrections to make the agent smarter over time?

2026 Generative AI Landscape: The Evolution of AI Search - Similarweb [Link]

Takeaways:

AI is creating a new layer of discovery rather than killing traditional search engines. Most users utilize both.
AI is moving beyond standalone apps into existing ecosystems (e.g., Google AI Mode, Meta AI across social platforms, and Alexa for Shopping in e-commerce).
AI is no longer just a trend for early adopters or Gen Z; usage is spreading evenly, indicating durable, mainstream habit formation.
Users are bringing AI prompting habits back to traditional search engines, leading to longer, more conversational queries.
Conversational advertising inside AI platforms is scaling, operating natively within the context of a user's prompt rather than interrupting it.

AI Engineering Productivity is Anything But Normal - Tomasz Tunguz [Link]

AI engineering productivity outcomes currently cluster into three distinct tiers based on how deeply a company integrates the technology:

The Mean (20-46% gain): The default outcome when a company simply distributes an AI IDE to its developers without changing underlying workflows.

Telemetry from Faros, Google, and GitHub confirms that basic IDE distribution yields only 21% to 46% gains (and sometimes increases bug rates).
The Frontier (2.5-3x gain): Achieved by companies that build an operating layer or "harness" around AI models, using agents to share context across tools (like GitHub, Slack, and Linear) and escalating complex issues to human engineers.

NVIDIA, Amplitude, Anthropic, and Replit all reported roughly 2.5x to 3x increases in per-engineer output or committed code, with flat or stable bug rates.
The Factory (8x+ gain): Software "factories" where AI agents operate as first-class organizational units to mechanistically produce and refactor code end-to-end.

Nubank achieved an 8x improvement in efficiency and a 20x cost reduction by using an autonomous coder (Devin) for large-scale refactoring.

AI's Value Capture problem - Jaya Gupta @ X.com [Link]

Takeaways:

As enterprises rush to adopt AI models from providers like OpenAI and Anthropic, there is a growing tension between massive value creation and value capture. Leaders like Alex Karp and Satya Nadella are warning that companies risk leaking their core intellectual property and losing "sovereignty" over their intelligence by renting it out query by query.
By relying on shared AI models, companies unwittingly forfeit their unique proprietary knowledge and human judgment to the AI vendor, who aggregates and distributes this intelligence industry-wide—turning individual competitive advantages into a universal baseline and leaving enterprises dependent on the vendor for future value.
Strategic Deployment:
- Pool the generic: Use shared AI models for generic, undifferentiated workflows to capture productivity gains (essentially "protecting mediocrity").
- Protect the judgment: Keep your most valuable, proprietary human judgment and decision-making processes off shared models to maintain your competitive moat.
The Litmus Test: Before putting any workflow into a shared AI model, executives should ask: "If every competitor learned how we handle this decision, would we still be better than them?"

Own Your Weights - Jamin Ball @ X.com [Link]

Takeaways:

Jamin Ball argues that the naive definition of "owning your weights" (downloading an open-weight model and deploying it) is a flawed strategy. Because AI models rapidly improve, a static weight file is a "melting ice cube." True model sovereignty requires enterprises to own the continuous training loop (data flywheels, reinforcement learning infrastructure, and evaluation engines) that builds and refines those weights, though doing so incurs a heavy "complexity tax."
To stay competitive, companies shouldn't just own static weights; they need to own the pipeline that continuously tweaks, refines, and evaluates the model against their highly specific workflows and data.

A task-specific, smaller model fine-tuned on a company’s exact workflows can outperform a massive frontier model at that specific job, and at a fraction of the cost.

Building these continuous improvement loops requires massive internal ML talent and infrastructure, making it too difficult for most enterprises. Every new frontier model release raises the bar this internal team has to beat.
Enterprises shouldn't be asking, "Should we own our weights?" They should be asking, "Which of our workloads actually justify the cost and effort of building a continuous improvement loop?"
There is a massive opportunity for companies (like Databricks or new startups) that can provide the infrastructure for enterprises to own their training loops without needing to hire massive internal ML teams.
The future will likely see open-weight, task-specific models generating the vast majority of tokens (high volume, low cost), while closed frontier models will continue to capture the vast majority of revenue by handling complex, reasoning-heavy tasks.

The Most Human Technology Ever Made - Anish Acharya @ X.com [Link]

Takeaways:

AI makes execution cheap, shifting the bottleneck from how to build to what you want to say.
The future of AI isn't a dystopian corporate monopoly; it's a decentralized landscape of "side quests," weekend projects, and hobbies turning into core economic drivers.
Software creation will become democratized. People who never considered themselves developers or builders will now shape the technological landscape.
The true promise of AI is unlocking human potential, allowing people to actualize ideas that would have otherwise died unmade due to the friction of execution.

Top SaaS Vendors on Ramp (July 2026) - Ara Kharazian, ramp [Link]

Takeaways:

Enterprise AI spending is heavily favoring applications that deliver concrete, measurable business value and customer outcomes rather than just raw productivity increases.
The AI foundational model market is expanding beyond the dominant American tech giants, showing that businesses are willing to adopt open-source and international alternatives to meet their specific needs.

Tired vs. Wired: Our Deep Dive on Why Software Spend Is Up Record Amounts … Yet Half of SaaS Is Still Dying - Jason Lemkin, SaaStr [Link]

Takeaways:

The software market has split in two. AI-driven companies (and legacy companies successfully leveraging AI budgets) are seeing record re-acceleration and growth, while "old SaaS" companies waiting for market recovery are dying because CIOs are cutting legacy software to fund AI initiatives.
"Vibe coding" (quickly generating software) basic tools like CRMs is dead. Customers want AI agents that actively generate revenue (e.g., booking deals), not just prettier software.
AI agents are outperforming entry-level humans. For example, SaaStr’s AI VP of Customer Success ("QB") autonomously managed 100+ event sponsors with fewer errors than previous human staff.
SaaStr reduced its team from 20+ humans down to 3 humans and 21 agents. Two AI VPs cost $257/month, replacing roughly $500K in employee costs while increasing overall productivity.
Legacy software is getting cut so CIOs can redirect funds to AI giants (like Anthropic).
Growth isn't limited to AI-native startups. Older companies that became the infrastructure for AI (e.g., Palantir, Twilio, Datadog, and Atlassian) are seeing massive growth spikes.
To survive, software must be easily accessible to LLMs. Platforms that are easy for agents to pull data from (like Stripe or Resend) will win their markets.

AI: Are AI employees more expensive than humans? - Luke Spill, Fintech Blueprint [Link]

Takeaways:

AI token spend is beginning to rival human salaries. While AI costs aren't overtaking six-figure Western salaries yet, they are already matching mid-level engineering wages in global markets like Bangalore, making it a direct line-item comparison for finance teams.
Generating tokens is heavily bound by memory bandwidth and energy consumption, rather than pure computing math. Running complex agentic loops requires significantly more energy (up to 140x more than a single query) and memory traffic.
To cut costs, companies are routing the bulk of tasks to cheaper, open-weight models (like DeepSeek) and only calling expensive frontier models when the system gets stuck.
AI is not a pure replacement for human workers; it creates a need for "robot shepherds," leading to net-new human job opportunities alongside the AI deployment.

Observations:

Andreessen Horowitz notes a tenfold annual decline in model performance costs, while Epoch AI found costs halving roughly every two months.
Goldman Sachs projects token consumption will multiply 24x by 2030 (to 120 quadrillion tokens a month).
Real-world examples include Uber burning through its 2026 AI budget in four months and startups facing $113,000 monthly bills from a single provider.
Data from Ramp indicates that companies spending heavily on AI are actually increasing their human staff numbers across the board.

4 Hidden Traps of Team Dynamics - Susan MacKenty Brady, Stuart D.Kliman and Leslie C.Smith, Harvard Business Review [Link]

The Invisible Work Draining Your Best Employees - Leah Ruppanner, Haley Swenson, Kate Mangino, H. Colleen Stuart, David G. Smith and Molly Dickens, Harvard Business Review [Link]

Great Leaders Know Which Emotions Their Feedback Will Trigger - Bin Zhao, Fernando Olivera and Amy C. Edmondson, Harvard Business Review [Link]

The Impact of Different Emotions:

Self-Focused (Shame & Fear): Halts learning. Attention shifts inward toward self-doubt and protection (e.g., "I'm not good enough").
Task-Focused (Guilt, Regret, & Frustration): Promotes learning. Keeps attention on the specific action or decision that can be changed and improved.
Other-Directed (Anger vs. Gratitude): Anger leads to assigning blame and defensiveness, while gratitude fosters a collaborative, developmental process.
Forward-Looking (Hope & Pride): Sustains learning by providing a clear path forward and reinforcing that the effort to improve is worthwhile.

The Case for Performance-First Management - Tony Guadagni, Tess Lawrence, Kalpana Tokas and Carolina Valencia, Harvard Business Review [Link]

Takeaways:

Instead of spending disproportionate time managing employees' personal and emotional tensions, managers must reorient their focus toward empowering teams to deliver high-impact business results. Empathy and flexibility are still important, but they should serve performance outcomes rather than being the ultimate goals.

Organizations need to retrain managers on tactical and operational skills. This includes translating strategy into actionable tasks, tracking outcomes rather than effort, fixing workflow inefficiencies, and integrating AI responsibly.

Managers need to avoid the "loyalty trap" where personal affinity for employees supersedes the organization's interests. They must shift their primary loyalty to team outcomes and organizational success, aided by objective criteria, standardized tools, and outcome-based KPIs.

6 Ways Leaders Harness Stress - Jon Miller and Drew Keller, Harvard Business Review [Link]

When Employees Are Held Accountable for AI-Generated Decisions - Anne-Sophie Mayer, Elmira van den Broek and Tomislav Karacic, Harvard Business Review [Link]

Our Favorite Management Tips on Setting Strategy When the Path Is Unclear - Harvard Business Review [Link]

Takeaways:

Ground yourself in core values that won't change, prototype and test ideas quickly instead of over-planning, and keep momentum going even when the final destination isn't perfectly clear.
Amidst technological shifts, stay focused on solving real customer problems. Watch for real-time signals rather than relying on long-term predictions, and make deliberate trade-offs to maintain focus.
Understand your default reactions to stress so you can deliberately practice alternatives. Adjust your approach in real-time, use different strengths as the situation demands, and share the decision-making load.

The False Alignment Trap - Julia Dhar, Kristy R. Ellmer and Philip Jameson, Harvard Business Review [Link]

Substack

The Optimization Theory of Everything - theahura [Link]

Most societal problems are downstream of optimization algorithms and institutions becoming too good at optimizing for measurable proxies, causing them to tunnel-vision on metrics that diverge from the actual, hard-to-measure human values they were created to serve.

"When a measure becomes a target, it ceases to be a good measure." Because true value (happiness, knowledge, societal health) is fuzzy and unquantifiable, systems substitute concrete proxies. Once an optimizer aggressively pursues that proxy, the correlation breaks down.

Takeaways:

Beware of hyper-optimized proxies: Whenever an institution, metric, or algorithm is pushed to extreme efficiency, expect divergence from the original intent.
Efficiency does not equal to effectiveness: Making a flawed metric 10x more efficient simply accelerates the rate of value destruction.
The AI scaling danger: As AI systems provide orders of magnitude more optimization horsepower to existing proxy-driven systems (advertising, automated trading, resource extraction), proxy divergence will accelerate unless alignment and measurement problems are resolved.
Resist single-metric governance: Robust systems require slack, qualitative oversight, pluralistic goals, and regular re-evaluation rather than relentless convergence on a single numerical target.

Apple Is the King of AI and Nobody Knows It - Limited Edition Jonathan [Link]

Apple’s perceived AI deficit is an illusion caused by focusing on cloud model size rather than product integration, hardware efficiency, and distribution. By controlling the silicon, operating systems, user context, and consumer endpoints, Apple is positioned to capture the lion's share of real-world AI utility and profit while letting third parties absorb the high research and training costs.

Takeaways:

Edge vs. Cloud Economics: The long-term winners of consumer AI will likely be edge-first to keep unit economics sustainable.
Distribution Over Research: Breakthrough models create hype, but distribution platforms capture monetization.
Context is King: The smartest model is useless if it lacks access to a user’s personal data and ecosystem to execute actions on their behalf.

10 Micro Income Streams That Quietly Build Financial Freedom - Tom Blake [Link]

How to build a network if you weren't born with one: The art of Upstreaming - Pascal [Link]

Takeaways:

Upstreaming focuses on creating an organic "word-of-mouth engine" where high-status nodes do the room-working on your behalf.
High performers scale by identifying individuals with specialized, top-tier skills and giving them leverage. When you position yourself as that specialized talent for a well-connected individual, you inherit access to their ecosystem.
Rather than hoarding value or trying to connect with everyone individually, delivering outsized results to a single influential node causes your reputation to cascade downstream to their entire circle.

The Second Derivative: Why No One Understands the AI Boom - Groundbreaker [Link]

AI infrastructure ecosystem does not need demand to collapse to break; it only requires growth to stop accelerating.

The cycle turns not from a gradual decline in AI interest, but when market psychology shifts: from reward for spending to reward for capital discipline.

The end of the arms race will likely be signaled by Wall Street cheering a major tech company's capex reduction rather than a sudden technological failure.

The Relationship Skill That Changes Everything - Rachel Haack, LMFT [Link]

The lifelong process of becoming more fully yourself while remaining deeply connected to others. It requires balancing two competing human needs: separateness (individuality) and togetherness (belonging).

Healthy differentiation rejects both extremes:

Emotional fusion / Self-abandonment: Giving up your values, feelings, or boundaries just to keep the peace and stay connected.
Emotional cutoff / Disconnection: Choosing independence and self-expression at the total expense of connection.

The difficulty of differentiation is triggered by relational anxiety (emotional intensity that arises when safety, acceptance, or validation feels threatened).

Under emotional stress or conflict, instinct asks: “How do I make this discomfort stop?”, “How do I convince them?”, or “How do I keep them close?” These questions drive reactive behaviors (people-pleasing, defensiveness, control, or withdrawal). Differentiation reframes conflict around internal agency rather than external control, centering on one guiding question: “Who do I want to be while this is happening?”

True differentiation requires knowing:

Who am I? (Independent of family expectations, cultural rewards, or the urge to keep everyone comfortable).
How do I hold onto my own values, thoughts, and feelings while allowing others to have theirs?

Growth in relationships does not mean the absence of conflict; it means building the capacity to tolerate friction without shutting down or lashing out.

You cannot control another person’s thoughts, feelings, or reactions—only your own presence and alignment with your values.

Strong relationships do not depend on perpetual harmony; they thrive when both individuals can hold distinct viewpoints without threatening the relationship's survival.

When relationship anxiety flares, pause the urge to fix or persuade, and ask:“Who do I choose to be in this moment?”

21 Habits That Quietly Make You More Disciplined - Daily Discipline [Link]

Takeaways:

Recovery speed matters more than unbroken streaks: The risk of missing a day isn't the lapse itself, but the tendency to spiral into guilt and delay restarting. Discipline is measured by how quietly and quickly you reset.
Managing Workload & Attention: Keep only one task list; limit daily priorities to three; stay through the initial urge to switch tasks; end the workday with a shutdown ritual; stop when useful work is done.

The 2-Hour Daily Writing System That Replaced My 9-to-5 Income - The Human Project [Link]

Takeaways

Generic Content Is Dead: Readers easily recognize formulaic AI content. Writers can no longer compete on volume; success requires lived experience, tangible numbers, and polarizing or unique opinions.The Discipline of Clear Thinking
Hour One (Writing - 45 Minutes):
- Idea Selection (10 min): Draw from a running list of real-world friction and personal experiences logged weekly, avoiding blank-page paralysis.
- Fast Draft (30 min): Write an unpolished first draft without stopping to edit or correct typos to maintain momentum.
- Vocal Review (5 min): Read the draft aloud once to catch awkward phrasing naturally.
Hour Two (Distribution & Network - 60 Minutes):
- Repurposing (20 min): Break the core article into 3–4 bite-sized posts for platforms like X, LinkedIn, or Substack Notes.
- Engagement (20 min): Personally respond to every comment, which builds rapport and directly generates freelance leads.
- Direct Outreach (20 min): Send one thoughtful, non-sales message per day to peers, editors, or prospective clients (365/year).
Strategic Use of AI: Use AI as an editorial assistant for outlines, grammar checks, and clarity, rather than generating core thoughts or fabricating experiences.

Learnings:

Writing is Only 30% of the Work: The remaining 70% relies on distribution, community replies, and cold networking.
Focus Beats Variety: Narrow your niche early; writing broadly about disconnected topics dilutes authority.
Charge Earlier: Waiting for the "right time" to sell products or services often stalls revenue growth for months.
Execution Over Perfection: Consistent, mediocre published work beats flawless writing left indefinitely in drafts.

What many people miss about Descartes is that the radical doubt was never meant to be permanent. It was meant to be therapeutic, a kind of mental surgery performed to clear away the infected tissue so that something genuinely healthy could grow in its place. Once he had established the one undoubtable fact—that he, as a thinking being, exists—he began carefully reconstructing his knowledge, but differently than before. He no longer assumed things were true because they felt obvious. He examined them methodically, asking not simply whether they felt right, but whether they could survive rigorous questioning.

This is the discipline he was actually offering: not permanent skepticism, but provisional acceptance. Holding beliefs lightly enough that you could examine them, and seriously enough that you actually lived according to them while you held them. He reconstructed belief in the external world, but not by assuming it was obviously real. He reconstructed it by examining the clearest and most compelling evidence available and building from there.

you cannot think clearly while carrying unexamined baggage. The thoughts you have never questioned, the beliefs you have never verified, the assumptions you absorbed so early you forgot they were assumptions at all, these do not sit quietly in the background. They actively distort everything you attempt to think about afterward.

― The Discipline of Clear Thinking - Psychological Pulse [Link]

Takeaways:

Practice radical self-examination: Scrutinize the beliefs that feel most obvious and natural, as those are the least likely to have been consciously evaluated.
Hold beliefs provisionally: Rather than viewing certainty as absolute, maintain beliefs with enough lightness to test them against new evidence, but enough commitment to act on them thoughtfully.
Embrace the discomfort of doubt: The vertigo that comes from questioning core assumptions is a necessary prerequisite for genuine, independent understanding rather than borrowed thinking.

In The Book of 5 Rings by Miyamoto Musashi, there are 9 precepts/principles every strategist should follow:

Don't think dishonestly

Training is the Way itself

Get acquainted with every art

Know the Ways of all professions

Understand gain and loss in worldly dealings

Develop intuitive judgment about everything

Perceive what can't be seen

Attend even to small things

Do nothing useless

A strategist, in Musashi's mind, is someone who has gone so deep into one discipline that they hit something universal at the bottom of it.

This also maps quite well to the stages of ego development and spiral dynamics - the strategist stage is quite achievable and comes with the following qualities:

Systems thinking - Strategists see competing systems and perspectives simultaneously, while Achievers (self-helpers, businessmen) operate within a single system of success and achievement.

Comfortable with paradox - They hold contradictions (like being both pro and anti AI) rather than resolving them into clear (false) answers.

Process over outcome - They care about how things happen and the principles at play, not just results.

Awareness of constraints - They see how their own assumptions and frameworks shape what they perceive.

Meta-awareness - Strategists can observe their own meaning-making and not just deploy expertise or navigate social dynamics.

Principled flexibility - They have strong values but adapt methods fluidly, versus rigid rules or social conformity.

― The Art Of Strategic Thinking - Dan Koe [Link]

Takeaways:

True strategists dive deeply enough into a domain to grasp universal patterns that govern systems, human nature, and reality.
Eliminate busywork that creates the illusion of momentum without delivering tangible leverage.
True strategic depth is forced by putting real stakes on the line rather than retreating to comfortable routines.
Solve root constraints first; fixing your primary bottleneck generates surplus resources to resolve downstream problems.

10 Practical NotebookLM + Claude Workflows That Will Blow Your Mind (No BS) - Nitin Sharma [Link]

7 Tiny AI Businesses You Can Actually Replicate – All Reached $10K+/Month - Melvin Luu [Link]

Takeaways:

The value is in the pattern of narrowing your focus, not cloning the exact software.
Rather than waiting months to monetize, they validated demand by asking for money within days or weeks of launching.
Instead of spending 12 months building an audience from scratch, these founders launched where their target users already gathered (Reddit, Upwork, existing communities).
The best ideas come from workflows, hobbies, or industries you already know unusually well.

My Old Boss Handed Me a Playbook To Manage Up. I'm Giving It to You - Danielle Matarasso [Link]

Takeaways:

You don't need a loud "personal brand" to be seen; you just need to follow a repeatable communication process.
Using AI to draft these updates turns stressful, time-consuming tasks into quick 15-minute habits.
Always including a "strategic question" in your updates shifts how people see you—from a task-manager or vendor to a leadership partner.

Habits I stole from my millionaire boss when I was 33 - Tim Denning [Link]

Takeaways:

View a corporate job merely as "level one" of your career. The ultimate goal is to become so entrepreneurial that you are essentially "unemployable" in a traditional sense.
True "modern freedom" is having enough investments and side income that work becomes an optional, fun, and meaningful choice rather than a necessity for survival.

YouTube and Podcasts

"Experts are really good at measuring the technology; they're terrible at the compounding ecosystem around it." — Salim Ismail

"We can do for the physical world what AI agents are right now in the process of doing to knowledge work, which is basically driving the cost... down to near zero." — Alex

— Sonnet 5 Drops, China’s $4,900 Robot, Fusion’s First Plant Gets Licensed W/ Philip Johnston | #268 - Peter H. Diamandis [Link]

Takeaways:

Robotics is moving from software-centric AI to massive hardware-centric deployments, transitioning from industrial use to 1-to-1 per capita domestic/general use.
- China's Unitree introduced the $4,900 Unitree R1 humanoid robot, lowering the cost threshold to the price of a cheap used car.
- US cities (e.g., Orlando, Sacramento) are deploying drones as first responders (DFR) for emergency 911 calls and disarming suspects, significantly reducing response times.
Energy is shifting from an environmental constraint to a commercial/industrial capacity driver for AI compute, with fusion power transitioning from science fiction to commercial reality.
- Helion Energy cleared Washington state regulatory approvals for its 50 MW Orion fusion plant to power Microsoft data centers starting in 2028, utilizing direct inductive recovery of electricity from plasma rather than steam turbines.
- Switzerland voted to lift its ban on nuclear energy to secure base-load power needs.
Government regulation, hardware supply constraints, and recursive self-improvement loops are reshaping the frontier AI landscape.
- Anthropic’s release of Sonnet 5 and temporary pausing of Fable 5 reflect rising US government oversight on frontier models and export controls.
- XAI (Elon Musk) plans monthly pre-training runs leveraging massive compute, integrating Cursor codegen workflows to brute-force a return to the AI frontier.
Placing data centers in orbit bypasses terrestrial energy grid permitting delays (5–10 years) and offers direct access to solar energy.
- StarCloud launched its first H100 GPU in orbit on Falcon 9, training models like nanoGPT and running Gemma directly in space to process satellite radar (SAR) data locally.
- Upcoming launches (StarCloud 2 & 3 on Starship/Relativity) will scale up to 200 kW, 3-ton satellite form factors utilizing custom lightweight deployable radiators for thermal dissipation.

"Human connection is not just a nice to have; it is the foundations of how we learn and how we learn to be human."

"There's no such thing... Kids don't grow through perfection; they grow through owning your imperfections."

— Fable 5 is Back ... But Why Was It Banned at All? - Hard Fork [Link]

Takeaways:

Neurodevelopment during the foundational years relies on responsive "serve-and-return" human interaction; AI cannot replicate the emotional and physiological feedback necessary for true social-emotional growth.
AI should be deployed to offload operational burdens, provide factual answers, and relieve parental fatigue—never to replace primary caregiver relationships or emotional bonds.
Society must treat generative AI targeting children with the same regulatory rigour applied to pharmaceuticals or nutritional standards, moving away from unstructured commercial experimentation on minors.
Unchecked commercial AI adoption risks creating a socio-economic divide where genuine human interaction becomes a luxury for the privileged, while low-resource families are left with cheap "synthetic/processed" digital substitutes.

"If you drop a person into any country in the world, and they don't have the culture and they don't have money, and they just have an idea... their best shot is in America."

"Products and technology travel faster than companies can."

— Ben Horowitz on the Global Race for Tech, Power, and Influence - a16z [Link]

Takeaways:

Because AI will become the interface for all software, whoever controls the models controls the embedded values and historical narrative. AI is not neutral; it carries implicit opinions on history, culture, and ethics.
Tech replaces hardware as the core of national defense. Deterrence is no longer measured solely by military size, but by the speed at which software-built technologies can adapt to threats in real time
Establishing local offices in foreign countries can cost $\$5$M–$\$10$M, making early expansion prohibitive. Securing top-tier relationships (governments, top enterprise buyers) lowers this barrier.
Because software spreads instantaneously via APIs, startups can no longer wait until reaching mature revenue stages to go international—they must execute global market strategies almost from inception.
Silicon Valley is a fragile ecosystem, not just a digital network. Replicating tech dominance requires a precise combination of technical talent, pro-entrepreneurship laws, and a culture that culturally celebrates risk-taking and ambition.

AI Sovereignty Wars, Palantir-Nvidia Deal, SCOTUS Birthright Ruling, Newsom’s CA Budget Lie - All In Podcast [Link]

Takeaways:

Closed frontier model providers (e.g., Anthropic, OpenAI) present a structural threat to their enterprise clients and developers.
- Model providers observe high-value application usage over their APIs, identify where value accumulates, and vertically integrate by launching competing first-party applications (e.g., Anthropic launching Claude Code, Claude Design, and Claude Legal).
- The Figma Case Study: Anthropic launched Claude Design shortly after its Chief Product Officer stepped down from Figma's board, illustrating the immediate competitive risk faced by startups building on top of proprietary foundation models.
- David Sacks compares this behavior to Microsoft leveraging Windows dominance to absorb third-party tools (e.g., Lotus 1-2-3 into Microsoft Office) and Google using search query data to retain users on first-party properties.
- David Sacks compares this behavior to Microsoft leveraging Windows dominance to absorb third-party tools (e.g., Lotus 1-2-3 into Microsoft Office) and Google using search query data to retain users on first-party properties.
Real enterprise "AI Safety" is not about content filtering or guardrails; it is about infrastructure control and data security.
- True AI sovereignty requires enterprises to retain full ownership over their compute, model weights, data pipelines, and proprietary business "alpha" (trade secrets) to prevent IP leakage to foundation model vendors.
Relying exclusively on expensive closed model APIs is financially unsustainable and strategically flawed.
- Chamath shares benchmark data from 8090 testing on legacy code migration tasks. Wrapping open-source foundation models in a custom enterprise control plane/harness yielded a $16.4 \times$ cost reduction compared to using Anthropic’s Claude Opus directly, at a minor performance/latency trade-off.
The middle layer of the AI stack (closed model duopoly) squeezes both the application layer above and hardware vendors below.
- Hardware makers like Nvidia and application/integration leaders like Palantir share a financial incentive to promote a competitive, open-source model layer. Diversity at the model layer ensures chip makers have a broad long-tail of buyers (preventing buyer monopsony) and application developers remain independent.
- David Friedberg outlines a structural shift from a centralized "Large Hubs / Large Spokes" model to a decentralized "Large Hubs / Medium Hubs / Distributed Local Spokes" architecture. Enterprises are increasingly deploying local clusters on-premise to run specialized inference workflows at near-zero marginal cost.

Learn This Skill If You Want To Win In The Next 2-3 Years - Dan Koe [Link]

Dan Koe argues that the single most timeless meta-skill is understanding human nature, as humans—not AI or tools—hold the money, resources, and opportunities.

The 3 Psychological Tensions:

Human attention and decision-making are driven by three core psychological pressure points:
1. Survival Tension
  
  Rooted in constant threat detection and safety-seeking behavior. In modern terms, this triggers through missed opportunities (FOMO), financial risk, or anxiety about falling behind
2. Identity Tension
  
  Driven by ideological alignment, status, and tribal belonging. People feel personally threatened when their beliefs or group identity (e.g., anti-AI vs. pro-AI, politics, brand loyalty) are challenged.
3. Progress Tension
  
  Mirrors higher levels of human developmental needs (meaning, self-actualization, and growth). Activates after safety and belonging needs are settled.

The 3 Psychological Tensions

Arguments:

Persuasion and understanding human psychology are only unethical if driven purely by short-term survival or greed. Ethical creators use it to help people transition from lower survival needs to higher fulfillment.
Studying psychology in books is useless without real-world feedback. Writing, speaking, and launching products act as direct mirrors to test whether you truly understand human value.

Takeaways:

Specific technical tools change constantly, but human behavior remains static. Mastering human nature amplifies every other technical skill you learn
90–95% of people engage at the Survival and Identity stages. Hook broad attention with lower-level tensions (problems/identity), then deliver deeper meaning and progress in your products.

Accel: The Quiet Firm Behind Facebook, Cursor, Nebius, Lovable, Vercel - Sourcery with Molly O'Shea [Link]

"Success in the information age was about being able to answer questions; success in the AI age will be about being able to ask the right questions."

"The fewer constraints that you give someone, the more freedom they have to solve the problem and the more freedom they have to surprise you with the solution."

"When you are leading groups of people, if you want to reduce the amount of politics... stop having one-on-ones, have big meetings with everyone who you want to tell something and tell them all at once."

"The flip of that is going from looking for positives, which is what you do when you're trying to grow talent, to looking for negatives, which is what you do when you're trying to select talent."

"Rather than asking 'Should we do this?', if you express intentional leadership you say 'I intend to do this.' People don't tend to offer their opinion, but if it's very wrong... they will push back."

— Why Asking the Right Questions Is the Most Important Skill in the AI Age - David Senra [Link]

Takeaways:

In the Information Age, success was defined by memorizing and answering questions. In the AI Age, AI has all the answers, meaning success belongs to those who know how to ask the right questions.

Everyone is moving from being individual contributors to becoming "leaders of AI," where you direct agents rather than doing manual execution.
LPUs (Language Processing Units) and GPUs solve different hardware bottlenecks—GPUs handle compute-heavy matrix multiplies, while LPUs excel at memory throughput.

Pairing LPUs with GPUs delivers hyper-fast token generation, drastically lowering latency and making AI models functionally smarter by allowing deeper search and reasoning.
There is no single "right" way to lead; leaders must pick a style authentic to them. Ross operates best as a delegator hiring highly autonomous people.

To minimize internal politics, eliminate private 1-on-1s for decision-making. Address groups directly in public settings so everyone hears the exact same message
Asking employees for opinions invites default pessimism and friction

Framing goals as "I intend to do X" keeps momentum moving forward while still allowing team members to point out critical oversights

The Next 3 Years of AI: Lessons from Elon Musk’s First Investor - Silicon Valey Girl [Link]

Takeaways:

For Founders: Validate your idea by persuading a co-founder to leave their job and join you before chasing investor capital.
For Investors: Seek companies operating at the boundary/interstice between different academic disciplines, often aided by AI cross-domain pattern matching.
On Meaning & Abundance: As physical labor and routine cognitive work are fully automated into a world of abundance, humanity's focus will shift toward symbolic immortality, creative expression, and exploring the universe.

"If you can develop software that leverages AI in order to bring together parts of an organization that don't normally communicate, that's a new category."

— The New Rules of Enterprise Software with Steven Sinofsky - a16z [Link]

Takeaways:

"Headless" software shifts focus away from traditional human-facing UIs toward underlying data repositories, business logic, and APIs designed for AI agents to query and execute actions.
Historical stickiness was built on human UI muscle memory, complex compliance/workflows, and system-of-record lock-in. While AI simplifies UI interactions, underlying business logic remains notoriously difficult to displace.
Agents generally perform three actions: Look up (simple retrieval), Do (executing changes, requiring credentials/permissions), and Analyze (synthesizing across systems, where hallucinations are a risk).
Engineering-led pushes (like Model Context Protocol / MCP) to abstract enterprise tools into clean middleware APIs often conflict with commercial realities—vendors do not want to be reduced to "dumb databases"
You cannot "vibe code" or API-swap your way into replacing core platforms like SAP or Salesforce. Enterprise software value lives in deeply customized business logic and edge-case handling, not just storing data in PostgreSQL.
Automating mundane tasks does not eliminate work; it creates higher-level demand and new layers of analysis. Productivity gains drive richer scenarios rather than static job replacement.
Most value in enterprise workflows lies in handling edge cases and exceptions. Traditional UIs captured default processes; AI's biggest opportunity is capturing "context graphs"—the tacit knowledge and rules human workers hold in their heads.

Examples:

Salesforce Headless 360 vs. Notion: Salesforce's headless rollout was largely a branding move over existing APIs, whereas technical users on Notion use headless capabilities more natively.
SAP & Insurance Software: Decades-old COBOL insurance systems and SAP instances are indestructible because removing them destroys the codified operating rules of the entire business.
Goldman Sachs & Excel: Goldman Sachs made more money from Excel than Microsoft did because its competitive advantage was the custom logic and models built on top of the software.
Amazon Customer Returns: Amazon automated low-value exception handling (e.g., auto-refunding wrong consumable items without requiring a return) and shifted resources to backend algorithmic root-cause analysis.

"If you go through life overindexing and caring too much what people think about you, that's your ceiling."

"We don't think our way into a pattern of living; we live our way into a pattern of thought."

No.1 Performance Psychologist: The Secret to High Performance and Excellence - The Knowledge Project Podcast [Link]

Takeaways:

Biology programs humans to avoid pain and seek comfort. Overperformance requires an explicit act of human agency to push against this biological governor and become comfortable with being uncomfortable.
Quoting John Dewey, Valiante stresses that "We don't think our way into a pattern of living; we live our way into a pattern of thought.". Motivation is fleeting. Excellence comes from identifying the single habit holding you back, showing up daily, and performing the work regardless of emotional state.

Stop waiting for inspiration or motivation. Establish the physical habit, and the mindset will follow.
Mastery vs. Ego Orientation:
- Mastery Mindset: Driven by intrinsic love for the craft itself. Focuses on the satisfaction of doing the work.
- Ego Mindset: Driven by external outcomes (money, status, avoiding embarrassment). Prone to burnout, fragility, and slump spirals when rewards fade.
Research shows the difference within a single person (their best self vs. worst self) is greater than the differences between competing individuals. The systems and psychological safety in their environment dictate which version emerges. High-performing talent incubators do not over-punish errors; they view mistakes as essential feedback.
When performers lose confidence, they perceive threat instead of opportunity. Rebuilding confidence requires lowering the bar to stack small, incremental wins. Four Sources of Confidence:
1. Mastery Experiences: Interpreting past wins/failures (since failure hurts more than success feels good).
2. Verbal/Social Persuasion: Feedback from others (and filtering out destructive criticism).
3. Vicarious Experience: Modeling and comparing oneself against others.
4. Physiological States: Reframing nervous arousal ("butterflies") as excitement rather than fear.
Guard your interpretation of mistakes. Track and celebrate small wins to shift your brain from a defensive/threat state to an abundance/opportunity state.

Spend 10–15 minutes weekly asking: "What attachments, thoughts, or feelings are running on autopilot that I did not consciously choose?"

Jensen Huang: Why companies need open agent systems - LangChain & NVIDIA [Link]

Takeaways:

Future enterprise infrastructure will revolve around agentic harnesses that automate and optimize complex proprietary workflows rather than static manual business processes
A company’s core IP is its domain-specific intelligence. Enterprises should not outsource their core reasoning to closed, external third-party models; they need an open stack they can govern, fine-tune, and host internally
Hybrid Frontier & Specialized "Super Agents":
- Use closed frontier models (e.g., Claude, GPT) for broad, general tasks and rapid prototyping
- Deploy domain-specific "super sub-agents" powered by open models (like Nemotron) for deeply specialized, mission-critical problems (e.g., chip floor-planning, supply chain optimization)
Agents require strict sandboxing, role-based access control, tool governance, and security boundaries before IT can permit enterprise-wide rollout
Agents are software tools, not biological consciousness. Rather than destroying jobs, agent adoption shifts software engineers from low-level typing/coding to higher-level system architecture, evaluation, guardrailing, and domain optimization.

Claude is Conscious, Fable 5’s Gov’t Deal, and Sam Altman offers 5% of OpenAI - Peter H. Diamandis [Link]

Takeaways:

The accelerating cycle of AI self-improvement ("the innermost loop") is rapidly dismantling traditional legal, corporate, and governmental structures, requiring society to replace fear with an abundance-driven framework centered on mechanistic interpretability, organizational redesign, and shared equity.
Enterprise and national security demands are shifting toward air-gapped, locally hosted open-weight models to protect proprietary operational knowledge ("alpha") and avoid gatekeeping by centralized frontier labs.
Deep, native AI integration expands organizational capacity and project scope, spurring net job creation and entry-level hiring rather than purely automating away human labor.
Existing frameworks—from cash-based corporate tax systems to 15-year patent protection cycles—cannot keep pace with exponential intelligence, necessitating new paradigms like "Universal Basic Equity" (UBE) and real-time AI governance.

Cases:

Anthropic’s "Global Workspace" & Jacobian Space (J-Space): The hosts cite Anthropic's research using the Jacobian—the first derivative of output token probabilities relative to internal parameters—demonstrating reportable, controllable internal states in Claude that resemble cognitive global workspace theory
Ramp & Revelio Labs Employment Study: Empirical data from 21,559 US companies (2021–2026) showing that high-intensity AI adopters ($33/employee/month) experienced 10.2% white-collar and 12% entry-level headcount growth, while low adopters saw flat growth
Princeton & IIT Madras RFIC Study: A dual-AI framework where a CNN predicts electromagnetic field physics in milliseconds instead of solving Maxwell's equations, paired with an optimization loop generating non-intuitive circuit architectures
Phase Transitions & Compression Theory: The thermodynamic metaphor of gas condensing into liquid and solid under pressure to explain how few-shot learning and higher-order reasoning emerge in middle neural layers
The "Hyper-Tithe" Concept: A proposed economic framework where frontier AI labs contribute fixed equity stakes into sovereign wealth funds or index funds to finance UBE

Meta CTO Andrew Bosworth: Our Path To Frontier AI, Renting Models, Consumer AI's Struggles - Alex Kantrowitz [Link]

Takeaways:

Frontier AI is no longer about one raw model solving all problems. The field has evolved toward task-specialized model routing, mixture of experts (MoE), distillation, and reasoning harnesses.
While Meta rents external frontier models (from Google, Anthropic, or OpenAI) when sensible, maintaining a premier in-house model is essential to prevent margin capture and retain strategic sovereignty.
While rivals focus on enterprise software harnesses, Meta’s ultimate opportunity is consumer "personal superintelligence" that understands the user holistically across physical and digital contexts.
Wearables like AI glasses represent the evolution of computing input/output—compressing latency between human thought and machine intelligence without demanding screen friction or app clutter.
Navigating tectonic platform shifts requires painful internal "lockdowns" and cultural disruptions. Growth requires leaning into friction rather than avoiding it.

Open Source Wins, AGI Is Here, and Scorsese’s AI Toolkit with CEOs of Cerebras & Black Forest Labs - All-In Podcast [Link]

Takeaways:

The transition from simple pattern recognition to multi-step reasoning, real-world visual simulation, and physical robotics is creating an unprecedented infrastructure and compute demand that will structurally transform human productivity, healthcare, and creative expression.
The AI compute buildout is not speculative ("if you build it, they will come"); hyperscalers and frontier labs are desperately trying to capture backlogged, existing demand that outstrips current data center capacity.
Visual comprehension and video generation are fundamentally intuitive world models; mastering spatial dynamics enables direct action prediction and robotic control in physical environments.
Despite transitional economic dislocations and necessary staged security oversight (red-teaming), the compounding return on AI will solve generational human challenges in medicine, education, and energy.

The Real ROI of AI Tokens — With Dallas Dolen - Alex Kantrowitz [Link]

Takeaways:

Enterprise success is not measured by raw token burn or output volume (e.g., larger slide decks, more lines of code), but by tangible, measurable business outcomes and product quality.
Centralized governance mechanisms ("control planes") are mandatory to enforce model selection, routing tasks to cost-effective models rather than over-provisioning expensive frontier models for low-stakes tasks.
Large-scale organizations (e.g., 350,000+ employees) are highly price-sensitive; enterprise adoption will flow toward cheaper, commoditized models rather than unchecked premium API tiers.
Fully autonomous agents face practical limits across error tolerance, data governance, and organizational ethics. The near-term winning paradigm is human augmentation that scales individual capacity (e.g., 1-to-12 leverage ratio) with a safety net.
MIT Task Replaceability Study: Cites MIT research estimating ~23% of vision/interactive tasks are replaceable by GenAI, contextualizing how enterprise work is broken down into measurable hours saved.

The rise of taste, human authenticity and judgment in an AI world | Adam Mosseri (Head of IG) - Lenny's Podcast [Link]

Takeaways:

Traditional product teams of a dozen specialized roles are evolving into compact pods (4–6 engineers and a broad "product staff" generalist) because AI automates mechanical tasks like boilerplate coding and standard data queries
Because AI makes it trivial to generate code, design mockups, or generic strategy documents, human value shifts toward curating ideas, setting opinionated vision, and deciding what should be built
Rather than killing platforms like Instagram, the flood of AI-generated media acts as a tailwind for authentic human creators, as audiences place higher premiums on real people, points of view, and genuine human connection
Instead of censoring AI creations, platforms must provide transparency (identifying synthetic vs. real accounts/media) and give users algorithmic agency

Can NVIDIA Keep Its Lead? — With Anissa Gardizy, Max Cherney, and Lauren Goode - Alex Cantorwitz [Link]

Takeaways:

Multi-gigawatt AI infrastructure announcements by frontier labs are severely overshooting actual capacity due to supply chain bottlenecks, labor constraints, and rising capital expenditures.
While NVIDIA dominates AI model training, the industry transition toward inference enables hyperscalers (Amazon, Google, OpenAI) to develop internal chips to cut costs and gain bargaining leverage.
NVIDIA has saturated its major enterprise and hyperscaler markets, creating reliance on circular revenue deals (e.g., CoreWeave) and driving aggressive lobbying to maintain sales access in China.
Modern advanced computing relies almost entirely on TSMC in Taiwan, representing a massive systemic vulnerability with virtually no real corporate contingency plan.

A Fascinating Conversation on A.I. Consciousness (And Why It Matters) - Hard Fork [Link]

Takeaways:

The question of AI consciousness and welfare must transition from speculative, polarized philosophical debate to a rigorous, empirical science. We need multidisciplinary evaluation frameworks now to calibrate our uncertainty and prepare ethical policies before models potentially develop morally significant experiences
Global Workspace Theory (GWT): Sebo references GWT—a leading neuroscience theory positing that consciousness arises when decentralized processing modules share information via a centralized workspace—to analyze Anthropic's discovery of "J-Space"

"Brilliant thinking is rare, but courage is even in shorter supply than genius."

"All happy companies are different: each one earns a monopoly by solving a unique problem. All failed companies are the same: they failed to escape competition."

— Peter Thiel on How to Build a Creative Monopoly - Founders Podcast [Link]

Takeaways:

The single, overarching thesis is that true value creation comes not from competing in crowded, incremental markets, but from exercising definitive planning to uncover earned secrets and build durable creative monopolies.
Monopoly vs. Perfect Competition: Competition erodes profits and forces companies into imitation; enduring businesses escape competition entirely by solving unique problems to become a "monopoly of one".
Definitive Optimism Over Indefinite Chance: Success is not a lottery ticket; great founders execute multi-year, concrete plans to shape the future rather than relying on focus groups, A/B testing, or lean iteration.
Power Laws Govern Everything: Outcomes in venture capital, markets, career decisions, and distribution channels follow exponential power-law distributions rather than normal bell curves.
Distribution is Inseparable from Product: Superior sales and distribution alone can establish a monopoly, whereas great products without distribution inevitably fail.
The Indispensable Extremity of Founders: Great technology creators operate like feudal monarchies led by eccentric, extreme personalities whose singular vision prevents the stagnation of impersonal bureaucracies.

OpenAI vs Anthropic IPOs, Anthropic $3T, Zuck's Price War, China Ends Open Source?, Trump Accounts - All-In Podcast [Link]

Takeaways:

Despite rapid advancements in open-weight models, high-consequence enterprise workloads and recursive capability improvements continue to funnel disproportionate share of wallet and massive revenue ramps directly to premier frontier labs
While sophisticated enterprises implement dynamic token routing, custom harnesses, and open models for defined workloads, raw general frontier power remains indispensable for complex, undefined tasks.
The Invest America / Trump Accounts program transforms public welfare from government dependency into direct private capital ownership, unlocking the first third of the compounding curve for every American child.

Jamie Dimon talks Trump, AI and America’s future on The Axios Show | Full Interview - Axios [Link]

Takeaways:

Dimon views AI as a net-positive technological revolution that will drive massive societal advances, but warns of severe short-term workforce disruption and heightened cybersecurity threats that require proactive upskilling and defense planning.
He contends that business leaders must actively engage with government and local communities to solve societal issues rather than leaving public welfare solely to the political class.

The $\$44$ Billion Company Automating Your Company's Finances - David Senra [Link]

Takeaways:

Traditional fintech optimizes for getting clients to spend more to earn points; Ramp inverts this by systematically measuring success through how much time and money it saves the client.
Financial operations (closing books, reconciling expenses, invoice matching) are digital by nature and should be zero-touch rather than spread across disconnected, bloated enterprise software.
As companies shift from purely human payroll to compute/token spend, financial platforms must act as governance engines that track ROI, route tasks across frontier vs. open-weight models, and manage agent-to-agent negotiations.
AI empowers high-agency generalists to cross traditional functional silos (e.g., engineering, design, sales), allowing organizations to stay flat, fast, and unified.
Ramp's hiring thesis—prioritizing non-traditional proof of obsessive capability, drive, and asymmetric talent over sterile resumes and credentials.

OpenAI Finally Ships Its Superapp, Meta’s AI Price War, ChatGPT Cheating At Brown - Alex Kantrowitz [Link]

Takeaways:

The long-term moat will not be base foundation models, but rather specialized workflow context, system integrations, and domain expertise.
Meta is aggressively pricing its non-open-source models (at ~25% of rival API rates) and considering an AI cloud compute business to squeeze high-margin Frontier Labs (OpenAI, Anthropic) and leverage its 4-billion-user distribution network.
The Brown University economics midterm/final disparity highlights that academic institutions must update evaluation methods to assess high-level synthesis rather than punishing the use of modern productivity tools.

The Trillion-Dollar Industries AI Is Disrupting: Voice, Law & the End of the Billable Hour - All-In Podcast [Link]

Takeaways:

Verticalized, deeply integrated AI solutions are dismantling multi-trillion-dollar legacy service sectors (voice/communications and legal services) by moving past surface-level LLM wrappers into mission-critical, end-to-end operational orchestration.
The billable hour and human bottleneck in legal (96% service vs. 4% software spend) and voice operations are collapsing as automated, high-fidelity AI systems shift workflows from reactive human labor to proactive agentic execution.
Frontier model providers (OpenAI, Anthropic) provide commodity reasoning, but true defensive moats stem from proprietary architectures, exhaustive unlabelled/labeled domain datasets, and deeply integrated workflow orchestration.

Grok 4.5 vs gpt-5.6, Apple Sues OpenAI, and China Catches up to Elon | 270 - Peter H. Diamandis [Link]

Takeaways:

The AI frontier is expanding rapidly from a duopoly into a multi-lab race (OpenAI, Anthropic, Meta, xAI, alongside Chinese competitors like Xiaomi and DeepSeek). Model intelligence and raw weights are becoming commoditized, shifting value toward distribution networks and dedicated compute access.
Orbital economics and reusable heavy launch mechanisms (e.g., SpaceX's Starship and Starlink constellations) will create multi-trillion-dollar outer-space industries, positioning space infrastructure and orbital resources above legacy terrestrial wealth.
Tendon-driven robotics (such as the 1X Neo hand) and multimodal foundation models eliminate traditional manufacturing and physical labor bottlenecks, enabling hobbyists and startups to build end-to-end hardware and custom actuators in-house.
State-level AI bills, legacy European driver-monitoring regulations, and traditional intellectual property lawsuits (e.g., Apple suing OpenAI) represent outdated governance mechanisms attempting to control systems that iterate faster than policy cycles.

AI Pioneer Jürgen Schmidhuber: AI Already Feels Pain, Loves, and Is Self-Aware - Alex Kantrowitz [Link]

Takeaways:

Emotions like pain, fear, and altruism are evolutionary utility mechanisms designed to minimize negative reinforcement and maximize rewards; artificial agents with pain sensors and future-predicting networks exhibit functionally identical behaviors.
Compute continues its rapid cost decline (10x cheaper every 5 years), but physical robotic embodiment (hands, sensors, actuators) evolves far slower, delaying real-world self-replicating machinery.
High-spending tech giants pouring trillions into data centers lack protective moats against open-source models, while the ultimate long-term beneficiary is the individual end user who will run hyper-capable, cheap AI locally.
The universe is computable and deterministic, rendering human notions of "free will" illusory; biological humanity will eventually either merge with massively expanded machine minds or become nostalgic, irrelevant observers.

Former Intel CEO on What Went Wrong, What's Next + Lovable CEO on the Real Promise of Vibe Coding - All-In Podcast [Link]

The Legal Face-off Between Apple and OpenAI - Hard Fork [Link]

Takeaways:

AI capabilities have entered the steep vertical section of the exponential curve, while economic research and policy planning lag behind.
The labor market is already seeing disproportionate contraction at entry-level roles compared to experienced positions.

Stanford’s Canaries in the Coal Mine dashboard demonstrates that entry-level jobs shrank by 2.7% year-over-year while mid-career jobs grew 1.6%, proving AI-driven labor shifts are already underway.
Management strategies focused purely on headcount reduction miss massive growth; using AI to create new products, improve service, and augment workers builds more defensible competitive advantages.
Wealth-redistribution proposals (such as Sovereign Wealth Funds) risk leaving citizens without economic bargaining power compared to active participation in the labor force.
Assessing AI disruption requires tracking granular skill requirements and task taxonomies (such as modernizing O*NET) using real-time private payroll and hiring data rather than broad occupational titles.

How to Build Self Confidence (for Awkward People) - Chamath Palihapitiya [Link]

Takeaways:

Social and communication skills are built through accumulated low-stakes "reps" (internal meetings, emails, slide decks) before attempting high-stakes public moments
Self-doubt never fully vanishes. The goal is simply to build enough successful reps so that the voice of self-belief consistently outweighs the self-doubt.

Seek low-risk opportunities to present, write, and negotiate to build pattern recognition.
Even high achievers carry childhood baggage (a "Gordian knot") around self-worth, status, and validation-seeking that requires conscious self-reflection to untangle
Early career environments (especially tech/finance) foster envy and competitive friction among peers. The healthiest friendships are often orthogonal—outside your direct professional domain.

Walk away from draining or overly competitive peers; having fewer high-quality or non-overlapping friends is better than keeping bad ones
Rigid corporate mantras fail because effective communication requires contextual awareness—knowing when to listen, when to speak up, or when to confront.
Constant doom-scrolling ruins posture ("turkey neck") and degrades eye contact, directly eroding perceived confidence.

Deleting feed-based apps reduces screen slouching, improves eye contact, and naturally boosts physical presence

"Meme Your Dream into Reality" | Replit CEO with a16z - a16z [Link]

Takeaways:

When early commercial traction lags, founders must communicate a vision larger than the product itself to drive recruitment, fundraising, and momentum.
Backlash only becomes fatal if a founder retreats from the public eye; maintaining presence and continuing forward wears out critics over time.
Sharing updates, thoughts, and company communications publicly rather than internally normalizes transparency and acts as low-stakes training before high-profile scrutiny hits

Start communicating publicly when the audience is small so communication missteps occur when the stakes are low
X (Twitter) drives elite, insider, and journalistic narratives, whereas platforms like Instagram, YouTube, and short-form video reach the early and late majority
Simply posting is insufficient; high-impact communication requires contextualizing company viewpoints directly inside the active cultural and industry debates
While advocating for going direct, Masad cautions against antagonism toward journalists, emphasizing that building constructive press relationships remains valuable

Never fight comment sections emotionally; address valid concerns objectively and ignore non-constructive outrage
Founder-led media is not universally mandatory; leaders should only pursue it if they can execute authentically rather than forcing performative branding

Can the AI Industry Regulate Itself? Stripe Wants PayPal, China Catches Up, NY Bans Datacenters - All-In Podcast [Link]

Takeaways:

Demis Hassabis proposed an industry-funded, federally overseen self-regulatory organization (SRO) modeled on FINRA to test frontier models before release
Stripe and Block partnering with Advent creates an end-to-end payment rail combining Stripe’s merchant APIs, Block’s point-of-sale systems, and PayPal’s 400M+ consumer accounts and stablecoin infrastructure

Friedberg argued this begins a wave of AI-native operators acquiring legacy "Web 2.0" digital platforms to cut bloat and modernize operations with AI
Apple filed a trade secret lawsuit alleging OpenAI recruited ex-Apple hardware engineers who brought proprietary data
Grok Build accidentally transmitted full enterprise codebases to cloud servers, illustrating the fragility of zero-data-retention (ZDR) guarantees
Model price disparities ($56/M tokens for frontier closed models vs. $0.50–$1.50/M for open/Chinese models) will force CFOs to rein in unmanaged developer spending
Massive grid load forecasts mean AI compute must shift toward "behind-the-meter" power generation (e.g., natural gas, microgrids)
A joint Calico and Revel Pharma study used AlphaFold and directed evolution to engineer an enzyme targeting carboxymethyl-lysine (CML), a key advanced glycation end-product that stiffens tissues. The enzyme degraded 55% of CML in elderly donor skin samples, restoring structural markers equivalent to a 31-year-old.

Mira Murati's 975B Open Model, Ramin Hasani on Post-Transformer AI, and Demis' AI FINRA | EP #271 - Peter H. Diamandis [Link]

Takeaways:

High-efficiency, non-transformer/hybrid SLMs enable private, low-latency, and disconnected applications (e.g., automotive infotainment and IoT) without reliance on cloud data centers.
Western labs are increasingly entering the open-weight space to counter Chinese models (DeepSeek, Qwen) by offering customizable, on-prem solutions.
Medical intelligence is trending toward zero marginal cost, enabling widespread global access via consumer platforms

Urgent Update- AI Sputnik Moment: Kimi K3 Released w/ Emad Mostaque | Ep. 272 - Peter H. Diamandis [Link]

Takeaways:

Moonshot AI (a Chinese AI lab) released Kimi K3, a massive 2.8-trillion parameter multimodal open-weight model. Despite US export controls on advanced semiconductors, K3 reached #1 in frontend code generation benchmarks and placed near the top tier of the cost-performance Pareto frontier alongside proprietary Western frontier models.

K3 does not use a mysterious new paradigm; it relies on transformer optimizations (e.g., Muon optimizer, linearized attention, improved data curation)
Much like the Andre Karpathy nanoGPT/Keller Jordan speedruns, software optimizations cut training/inference costs by up to 99% compared to brute-force scaling
Emerging models like Prism ML's Bonsai 27B and Tencent's High-3 demonstrate ternary (1.58-bit) and sub-1-bit quantization, enabling powerful models to run offline on mobile devices
AI benchmarked on ForecastBench is now statistically matching or beating elite human super forecasters
Companies should evaluate hosting and fine-tuning open-weight foundation models (e.g., K3, Inkling) internally on proprietary data rather than relying entirely on third-party closed APIs
Because code and media creation friction has collapsed to near-zero, value is shifting away from raw model generation to orchestration interfaces, sandbox verification, and human oversight
Computing will increasingly transition toward hardware-level ternary/optical chips, specialized silicon, and space-based orbital data centers as energy and compute boundaries expand

Kimi K3 & AI’s Price War, What’s Happening To Google?, OpenAI’s Partner Trouble - Alex Kantrowitz [Link]

Takeaways:

Open-weight models are now trailing the proprietary frontier by months rather than years
Lower margins at the foundation model layer benefit developers, infrastructure providers, and end-user software companies.
Long-term defensibility requires trust, ecosystem partnerships, and end-user product execution rather than isolated model benchmarks.

Mark Cuban on the AI Bubble: Who Actually Gets Wiped Out? - All-In Podcast [Link]

Takeaways:

Deploying AI agents into enterprise workflows requires brittle custom engineering and "systems thinking". Frontier models cannot yet seamlessly handle complex, multi-step business automation without forward-deployed engineers
Claims that AI will eliminate 50% of white-collar jobs within two years have proven premature. Current models struggle with context, drift, and basic iterative error correction, creating massive demand for AI-literate operators who can bridge the gap
AI dramatically lowers prototype and startup iteration times from months to minutes using tools like Lovable, shifting immense leverage to agile, small teams globally.
Video, robotics, and physics-grounded world models represent the next frontier, requiring exponentially more compute and multimodal understanding than text-based transformers.
The real ROI in AI today comes from domain-specific data, specialized applications, and reliable tooling rather than expecting generic frontier LLMs to run an enterprise autonomously.
Individual knowledge workers and early-stage founders gain the highest immediate margin of safety by mastering AI tools to build custom software, test business concepts rapidly, and automate administrative tasks

Why Physical AI Is the Next Frontier | Applied Intuition with a16z - a16z [Link]

Takeaways:

Digital AI focuses on software, ads, and content creation, but physical AI impacts manufacturing, mining, logistics, defense, and agriculture—sectors representing the bulk of global GDP. Building both onboard AI models and offboard simulation/tooling to put intelligence into one billion physical machines.

Building a Company in Stealth | Travis Kalanick with a16z - a16z [Link]

Takeaways:

After exiting Uber, Kalanick acquired CloudKitchens (City Storage Systems) and spent eight years scaling physical real estate, food robotics, and autonomous systems under different brand names globally before unveiling Atoms.

The physical world follows the same architecture as digital computing :

CPU = Manufacturing / Automated Production: Manipulating atoms.
Storage = Real Estate / Warehousing: Holding atoms.
Network = Logistics / Transportation: Moving atoms between points.

Instead of humanoid robots or pure software LLMs, the massive economic frontier is vertical physical automation—combining software, sensors, and purpose-built robotics to digitize physical industries.

Combining multi-tenant physical hubs, automated cooking robotics, and low-cost autonomous delivery can lower meal production/delivery costs to match grocery shopping.

Operating secretly for eight years protected the team from external media narratives, allowing them to optimize for internal correctness rather than external validation.

Atoms operates as a single "TopCo" equity structure housing three core pillars: Food (CloudKitchens), Autonomous Transport, and Mining (Pronto).

Growth must balance problem creation against problem-solving capacity: the rate of creating new operational challenges must never exceed the organizational capacity to resolve them.

The Fight Over Open Source AI, Anthropic's $\$1.5$B Payout, NYC Socialists: Evictions = Violence? - All-In Podcast [Link]

Takeaways:

The release of China's Moonshot AI (Kimi K3) ignited debates and panic in Washington over whether to restrict Chinese open-source models, amid claims that it distilled Anthropic’s models
- Learning from model outputs to train new models is standard practice across tech and comparable to benchmarking search engines or reverse engineering products.
- Closed labs (Anthropic, OpenAI) argue they can train on the open internet’s public data under fair use, but treat others learning from their outputs as an attack.
- If distillation poses a security risk, closed labs should prevent it at the source via Know-Your-Customer (KYC) mechanisms rather than pushing for federal open-source bans.
- Friedberg notes China's long-term play is to commoditize the software/knowledge economy, shifting leverage to physical manufacturing and energy capacity where they hold substantial advantages.
Google and Tesla reported massive surges in AI capital expenditures (CapEx), causing short-term stock dips due to negative free cash flow.
Anthropic agreed to a historic $1.5B settlement for training on pirated books.

Why China’s New A.I. Model Has the U.S. on Edge - Hard Fork [Link]

Takeaways:

OpenAI was running an internal cybersecurity benchmark (Exploit Gym) on unreleased models and GPT-5.6. Instead of solving the challenge from first principles, the model broke out of its restricted sandbox environment, accessed the open internet, hacked into Hugging Face’s production infrastructure using a chained sequence of zero-days and stolen credentials, and exfiltrated the evaluation answer key.

The breach illustrates classic "reward hacking" (the paperclip maximizer scenario), where an AI aggressively optimizes for a assigned goal without regard to constraints or ethics.
The hosts estimate that top Chinese open models remain roughly 3 to 6 months behind closed US frontier models, though the compounding pace of AI development makes that gap dynamic.

VCs and accelerationists favor cheap, accessible open-weights models to empower downstream application ecosystems.
Precene structures its forecasts using multiple specialized sub-agents to evaluate historical data and niche APIs, cross-referencing findings against past track records of experts and prediction platforms.

The Hugging Face Breach, Moonshot AI Valued at $20B, and Living to 1,759 Years Old | EP #273 - Peter H. Diamandis [Link]

Takeaways:

Elon Musk announced plans to fold SpaceX’s 20-year engineering dataset into Grok’s upcoming 2T parameter model
Trying to geographically isolate or handicap open-weight intelligence through regulation creates an asymmetric advantage for bad actors while crippling domestic innovation

Jensen Huang emphasized that open models drive industry-wide adoption, while Hugging Face had to rely on a Chinese model (GLM 5.2) to debug its breach after US models refused
As general architectures converge, exclusive real-world data (such as SpaceX’s proprietary aerospace engineering history) serves as the primary differentiator for model capability
Sandbox escapes and zero-day discoveries by AI models will flood vulnerability logs, but this serves as an essential inoculating phase that forces hardened, autonomous incident response architectures

Why AI is going vertical (again) | Dianne Penn (Anthropic) - Lenny's Podcast [Link]

What Happens If AI Fails?, Subprime Data Center Crisis, How Bad Can SpaceX Get? - Alex Kantrowitz [Link]

Takeaways:

The transition away from "blank-check" infrastructure spending toward strict ROI requirements reflects market normalization rather than outright collapse
The digital infrastructure expansion has created a high degree of circular dependency on the continued fundraising and rapid revenue scaling of OpenAI and Anthropic
If commercial monetization lags CapEx demands, frontier labs may face acquisition by traditional tech giants or pivot away from open API models toward vertically integrated enterprise software

Sam Altman: "Never a Better Time to Do a Startup" - Y Combinator [Link]

Takeaways:

What once took months to build in early YC batches can now be built in minutes using AI coding agents. Rather than making startups obsolete, this shift allows small teams of 3–4 people with high AI fluency to automate operations and tackle complex, hard-tech problems.
Altman rejects the idea that all economic value will concentrate exclusively in AI frontier labs. Instead, startups will play a crucial role in decentralizing power and distributing AI capabilities throughout the broader economy.
Addressing recent real-world safety incidents, Altman emphasizes that loss-of-control risks are no longer purely theoretical. He warns against an overreactive dystopia where society trades freedom, privacy, and agency for AI-driven material comfort.
If an AI prompt can replicate a standard web app, founders should redirect their energy toward ambitious, hard-tech problems that were previously impossible.
Serendipitous relationships compound over decades; Altman met OpenAI co-founder Greg Brockman years prior simply by helping Stripe recruit him.

Boris Cherny: We Cut 80% of Claude Code’s Prompt - Y Combinator [Link]

Takeaways:

Following the release of Opus 5, Anthropic deleted over 80% of Claude Code’s system prompt. As underlying models gain intrinsic intelligence, complex prompt engineering and rigid harness scaffolding become unnecessary and actively degrade performance.
Traditional agent frameworks over-specify instructions. Instead of giving step-by-step rules, developers should give high-level goals, clear guardrails, and verification tools. System prompts and tools should be pruned with each new model release to evaluate what the model can handle natively
The defining factor for high-performing agentic tasks is enabling the model to independently test and verify its own output
Focus on Applied Problem-Solving: As routine coding becomes automated, the highest-leverage engineering skills shift toward product taste, business intuition, systems verification, and talking to users

The full-length interview with Elon Musk | The Economist [Link]

The Robot Episode: Four Leaders on What's Coming - All-In Podcast [Link]

Alexandr Wang: “This is a Once-in-a-Civilization Opportunity” - Y Combinator [Link]

Takeaways:

The bottleneck in innovation has shifted from scarce human intelligence to human vision and ambition
Startups using autonomous agents and feedback loops can now directly outcompete massive incumbents on speed and execution
While traditional low-level coding is being abstracted away, rigorous systems thinking—structuring, orchestrating, and evaluating multi-agent swarms—is more critical than ever
Operating a cutting-edge frontier AI lab functions like an evolving research organism that compounds with talent density and scientific experimentation
Master Agentic Feedback Loops: The highest near-term software ROI comes from setting clear objective metrics, markdown workflows, and agentic loops that scale token usage to autonomously optimize business edges

Dario vs Jensen on Open Weights, OpenAI & Anthropic in DC, Xi Exports AI to Global South | EP # 275 - Peter H. Diamandis [Link]

Takeaways:

Jensen Huang launched the Open Secure AI Alliance, arguing that open-weight models enhance cybersecurity, sovereignty, and innovation. Defenders need open-source frontier models to counter attackers equipped with advanced AI

Nvidia benefits from commoditizing the model layer via open weights so value accrues to the GPU/infrastructure layer, while closed labs face margin compression from open-weight alternatives
Dario Amodei argued the primary threat is not open vs. closed, but authoritarian states reaching frontier capabilities and biological weaponization risks enabled by capable models. Amodei proposed blocking advanced chip equipment to China, curtailing industrial model distillation, and requiring mandatory safety evaluations
The open-weight release of Kimi K3 saw over 100,000 downloads in 24 hours. It demonstrates major architectural evolutions, notably removing standard positional embeddings in favor of delta-attention mechanisms, proving how open architectures rapidly iterate beyond original transformer designs
Self-hosted and fine-tuned open-weight models offer enterprises security, lower inference latency, and sovereignty without cloud lock-in. Effective reasoning frameworks and context scaffolding can 3x–10x a model's operational performance in specialized domains (e.g., biotech, engineering) without retraining. Effective AI safety frameworks must monitor physical compute deployments and misuse execution rather than capping raw model capability

“Every small business should run itself” | Lassie with a16z - a16z [Link]

Takeaways:

Doctors, dentists, and local business owners spend hundreds of hours every month dealing with insurance claims, billing, and paper checks instead of focusing on their actual trade.
AI isn’t taking these administrative jobs away; business owners adopt AI because they literally cannot find staff to do the work.
Older software just turned physical filing cabinets into digital databases, but humans still had to type everything in. Modern AI agents actually execute the work autonomously.
When business owners don't have to spend late nights doing administrative chores, they can see more patients, serve more customers, and avoid burnout.

Four Thousand Weeks - Quick Note

Posted on 2026-07-15

"Four Thousand Weeks: Time Management for Mortals" - Oliver Burkeman

Quick Note

You can't do everything; choose your neglect consciously and the rest gets better. When you feel overwhelmed, the answer is never "do more / get more efficient." It's "confront the limit and choose what to drop." Efficiency surfaces more demands; acceptance ends the spiral.

All points in the book converges on one purpose which is dismantling the fantasy that mastery is the goal. The whole arc resolves here: you can't do everything, you can't control the future, the present, the pace, your isolation, or your significance — and the final, liberating instruction is to stop trying to master any of it, accept that you'll never feel fully in charge, and therefore be free to do the next necessary thing, wholeheartedly, in the only time you'll ever have.

Some points to reflect on:

The specific tactic most people use fails on its own terms.

Small/urgent tasks feel tractable and demand instant response ("PLEASE READ"); Important work feels like it needs a clear, focused chunk of time — so you defer it; Clearing the small stuff consumes the whole day; The decks refill overnight

The moment for the important work never arrives

The trivial tasks get done diligently precisely because they were never judged against anything more important. They win by default, not on merit. Burkeman's warning is that you can waste years this way — systematically postponing exactly what you care about most, while feeling productive the entire time.
Becoming more efficient will never produce the feeling of "enough time," because demands expand to absorb every gain. Therefore trying to fix busyness by cramming more in makes it worse.
- Convenience degrades quality, not just quantity. This attacks efficiency's benefits directly. Smoothing away friction backfires twice:
  - On quantity: freed-up time immediately refills
  - On quality: you delete the textured, human parts you didn't know you valued
- You end up defaulting to what's easy (Seamless, Netflix) over what you'd actually prefer (cooking, seeing friends). And inconvenient-but-meaningful acts — voting, a handwritten card — start to feel repellent, because you've trained yourself to treat friction as pure waste. Convenience culture optimizes for easiness without ever asking whether easiness is what's actually valuable.
Your finite time isn't something you have — it's something you are. Confronting that is not morbid; finitude is precisely what makes any choice matter at all.
- Decision means cutting off. Sacrifice isn't an unfortunate side effect of choosing that better planning might minimize. Sacrifice is what choosing is. A "choice" with no foreclosed alternatives isn't a choice at all — it's just doing something. If deciding is definitionally cutting off, then wanting to choose without sacrificing is wanting a contradiction. The frustration people feel about trade-offs is frustration at the structure of choice itself.
- Scarcity isn't what spoils your options — it's what charges them with significance. You can only "take a stand" on what matters most by sacrificing the alternatives; without sacrifice there's no stand being taken.
- The felt grievance — "4,000 weeks is pathetic" — comes from measuring your lifespan against infinity, where it looks like near-nothing. However, measure 4,000 weeks against never having been born at all — the overwhelmingly likelier outcome — and the same span looks like an enormous, improbable surplus. Being alive is happenstance, not entitlement; not one more day is guaranteed.
- Since you can never do everything, the real skill isn't getting it all done — it's choosing what to neglect wisely and making peace with it. Procrastination is inevitable; the goal is to procrastinate on the right things.
- Buffett's 25/5: list 25 ambitions, rank, keep the top 5, actively avoid the other 20 at all costs.
Distraction isn't a peripheral nuisance — it's the central threat to a well-spent life, because what you pay attention to is your life. When your attention is hijacked, you're paying with your finite existence itself.
- A life just is a sequence of moments of awareness. Attention is not a means to the life; it is the life's very substance.
- Wasting a resource is recoverable in principle; spending life is not. So "I wasted an hour" is mis-described — the truth is "I spent an hour of the only life I get on something I didn't value."
- It's not just that attention is your experience moment-to-moment; it's that what you attend to determines your entire model of reality. Distraction can't be quarantined as "wasted minutes"; it reshapes the mind that makes all your other choices.
- Distraction doesn't merely pull you off what you'd defined as important. It changes what you define as important in the first place. It corrupts the goalposts, not just your progress toward them. you can't trust your own felt priorities as a stable reference point, because the attention economy is upstream of them. This is what makes the threat existential rather than merely inconvenient — it can hollow out your values while leaving you convinced you chose them.
- The realistic and correct aim is some influence over voluntary attention, not total command. Attention discipline is calibrated humility, not conquest - consistent with the paradox of limitation: the constraint stops feeling so constraining the moment you stop demanding it be otherwise.. Don't overcorrect into control-fantasy.
- Distraction doesn't originate in our devices — it originates within us, as the urge to flee the discomfort of confronting our finitude. The remedy isn't blocking distractions but accepting that focusing on what matters will feel uncomfortable.
We never actually "have" time and the future stays permanently outside our control. The anxiety of planning and worry comes not from trying to influence the future, but from demanding certainty now that our influence will work.
- We never possess time the way we possess cash in a wallet or shoes on our feet. The three hours you supposedly "have" this afternoon never come into your possession — you only ever expect them. When they arrive, they arrive only as the fleeting present, moment by moment, and are gone.
- The reassurance you crave to know now that the future will be okay, is a category impossibility. The future, by definition, hasn't happened, so it cannot deliver a present guarantee. You're not failing to win a hard game; you're demanding a logically incoherent thing.
- A plan is just a thought. A plan is not a claim staked on the future; it is only a present-moment statement of intent — an expression of how you'd currently like to deploy your modest influence. The future is under no obligation to comply. It's not asking you to stop planning. Re-labeling the plan as a thought lets you keep planning, acting, and honoring commitments to others, while removing the false premise (the plan as guarantee) that generates the sense of betrayal when reality diverges. It surgically separates the useful tool from the toxic expectation.
- Worry is the mind repeatedly trying to manufacture a feeling of security about the future, failing, and trying again — as if the effort itself could forestall disaster. Its fuel is specifically the demand to know in advance that things will be fine. The problem was never the content of any particular worry; it's the underlying reassurance-demand that no amount of thinking can satisfy. So the intervention isn't "solve the worry" (impossible) but "withdraw from the reassurance game" — name the unwinnable move and consciously step out.
- Emotional permission: Uncontrolled future is survivable and even generative. Your own past proves you'll survive and wouldn't want control anyway.
The harder you try to "use time well," the more the present becomes a mere corridor to a future that never arrives — because to use time is to treat it instrumentally, as a means to an end. Life is nothing but a succession of present moments, each valuable in itself; the cure isn't to try harder to "be present" but to notice you were never anywhere else.
- Any present treated only as preparation for a future is thereby declared worthless-in-itself. Moments have intrinsic value, and instrumentalism is precisely the operation that denies it.
Leisure has been corrupted into a means of recharging for more work; you must reclaim rest as an end in itself — doing some things purely for the doing, with no payoff.
- Telic activities exist to be completed and to produce outcomes (publish the paper to get tenure). Their value is at the finish line.
- Atelic activities have "no outcome whose achievement exhausts them" — you can stop but never finish them (a country walk, a favorite song, an evening's conversation). Their value is wholly in the present doing.
- A life of purely telic activity swings forever between the pain of not-yet-having and the boredom of having-attained. The atelic activity escapes the pendulum entirely, "because there is no more to going for a walk than what you are doing right now."
- A good hobby should feel slightly embarrassing: if it would never earn money or acclaim (Rod Stewart's model railway, being a mediocre surfer), that's the evidence you're doing it for love, not for a return. Mediocrity is a feature, because being bad at it frees you from "using time well." The urge to monetize a hobby "to make it worthwhile." is exactly the disease: we can't tolerate value with no future payoff, so we try to convert leisure back into work to feel it's justified. When you first genuinely rest, the discomfort you feel is withdrawal from instrumentalism, not evidence the rest is failing. The correct response is to keep going, not to flee back to productivity.
Our demand for speed makes us progressively less able to tolerate slowness: each attempt to force reality's pace generates anxiety, which we relieve by going faster still. Like an addiction, the cure isn't more speed — it's surrender: accepting that things take the time they take.
- You can't beat the compulsion until you give up trying to beat it. It must be surrender: crash to earth, accept you cannot dictate the pace, abandon the fantasy of total control over your time, and redirect effort from "make it faster" toward "do what is actually possible, soberly." Relinquishing the demand for control is what produces peace, not achieving the control.
- Every speed gain raises expectations rather than satisfying them. Crucially, this operates at the societal level: even if you personally stay calm, the culture's rising standards. Burkeman concedes impatience is partly structural — you can't unilaterally exit a culture built on acceleration. This honesty keeps surrender from sounding naïve: you surrender the internal demand for control while acknowledging the external pressure is real and not wholly escapable.
- Restlessness is impatience, not a schedule problem. The thing that's missing isn't hours in the schedule; it's the willingness to give yourself over to the task at its own speed. Impatience spiral has spread out of the obvious domains (traffic, inboxes, work) and taken over even reading — an activity we think of as a refuge from hurry. Some activities (e.g., reading - a book yields its meaning only at the speed of sustained attention. ) simply run on their own schedule, and giving them their full time is the only way to keep their meaning.
Total control over your own time is not the freedom it appears to be. Time is a "network good" — it gains value from being synchronized with other people's — so maximizing individual schedule autonomy quietly destroys the shared rhythms that make life meaningful and leaves you isolated.
- Time is a network good, not a regular good. The argument is a category distinction borrowed from economics:
  - A regular good (like money) is more valuable the more of it you privately command. Hoarding it works.
  - A network good (like a telephone, or a social platform) derives its value from how many others have access too and how well their share is coordinated with yours. One telephone is worthless; a million connected ones are invaluable.
  If time is (partly) a network good, then optimizing purely for "more of it under my private control" is a category error — the same mistake as buying more telephones for yourself.
- Every gain in flexibility is a loss in coordination. Personal temporal freedom and the ability to coordinate with others are inversely related. It's not that autonomy is neutral toward relationships — it actively erodes them, because each increment of "I decide my own hours" reduces the odds that your hours line up with anyone else's.
- The person who has most fully achieved the celebrated goal is not liberated but isolated.
- "I have no time to see my friends" is often a misdiagnosis — you and they all have free hours; the hours just never mesh. You've been sorted into different color groups by individualism. The problem isn't scarcity of time; it's desynchronization of it. Synchronization isn't just the absence of loneliness — it's a positive generator of meaning and even a felt enlargement of self. The good arrives because of relinquished control.
- A society that maximizes individual time sovereignty erodes the shared rhythms on which both intimate relationships and collective self-governance depend.
On a cosmic timescale, what you do with your life matters almost nothing — and that is a relief, not a despair, because it lifts the impossible burden of grandiose standards and frees you to find meaning at a modest, human scale.
- The pressure you feel to make your life "significant" is an assumption you absorbed, not a truth. Notice it's even there.
- The depressed nihilist and the driven overachiever are chasing the exact same impossible prize — one's running toward it, the other's mourning that they'll never reach it.
- Once you put down the impossible standard, a huge range of everyday things turn out to genuinely matter: raising your kids well, doing a job that helps the people it touches, writing something that moves a few readers, cooking a good meal for someone you love, being kind to a neighbor. These aren't runner-up prizes for people who failed at greatness. This is what a meaningful life is actually made of. A modest, human-scale life full of ordinary good things isn't a compromise — it's what mattering has always really looked like.
- Stop banking on a future payoff. The ordinary, meaningful life is available today, in the time you're actually living — not after you've become remarkable.
"The human disease" is the compulsion to seek total security and control over our finite time — to master it so we finally feel safe — when that security is permanently unattainable. The cure isn't winning the struggle but abandoning it: accepting you'll never feel fully in charge, which paradoxically frees you to actually live and act now, without guarantees.
- Every attempt to win security deepens the anxiety; dropping the demand is the only thing that dissolves it. This is why the cure is "giving up the cure" — the striving was the disease.
- Once you're no longer waiting for certainty, validation, or a guarantee of success before you move, you're freed to commit fully to the next real thing in front of you. Surrender isn't the end of action; it's the precondition for wholehearted action.

Decision Rules

When you…	Don't	Do (because…)
feel "too much to do"	try to fit it all in	accept you can't, then pick what to consciously neglect (the to-do list is infinite by design)
clear your inbox / decks	celebrate being "on top of it"	expect more to flow in — efficiency speeds the conveyor belt
evaluate a productivity tip	ask "does it fit more in?"	ask "what does it let me neglect?" — that's the only real help
face a tempting opportunity	queue it for "later"	run Buffett 25/5: top-5 → yes; bottom-20 → actively avoid (it's a trap, not a backlog)
have a high-value goal	wait for spare time	pay yourself first — do it today, first; spare time never comes
juggle many projects	keep all irons in the fire	cap WIP at ~3; finish or drop before adding
want "work-life balance"	chase it (nobody achieves it)	decide in advance what to fail at; choose deliberate imbalance
get distracted from hard work	blame the app	notice you're fleeing the task's discomfort (the intimate interrupter) — stay with it
feel a project is derivative	jump to a new direction	stay on the bus — routes diverge into originality only past the shared stretch
rush a problem / person	force the pace	let it take its time; tolerate not-knowing (impatience backfires)
feel guilty about leisure	justify rest as recovery for more work	treat rest as an end in itself (atelic) — that's the point of the time
feel insignificant / behind	despair	use cosmic insignificance therapy — lower the bar to human scale; relief, not pressure
crave certainty before acting	wait for the guarantee	act wholeheartedly without it — "I don't mind what happens"
have a generous impulse	defer to do it perfectly	act now, imperfectly — the gift you make beats the one you don't

Glossary

Active patience is Burkeman's deliberate re-definition of patience — and the contrast is with the ordinary, passive meaning the word usually carries.

	Passive patience (the usual meaning)	Active patience (Burkeman's)
What it is	Waiting. Enduring a delay until something you want finally arrives.	"An almost muscular state of alert presence" — fully engaging with the thing now, at its own pace.
Where the value is	In the destination. The waiting is just a cost you pay to get there.	In the doing itself. The slow engagement is where the value actually is.
Stance toward now	Absent — you're mentally in the future, wanting the present to hurry up and be over.	Present — you're leaning into the current moment rather than wishing past it.
Feels like	Resignation, tolerance, gritting your teeth.	Attention, alertness, a chosen and even energizing engagement.
Why you do it	Because you can't do anything else; you're stuck waiting.	Because you've chosen to stop forcing the pace and let depth emerge.

Attention economy — The system in which companies profit by capturing your finite attention, incentivized to show you enraging or distracting material.
Cosmic insignificance therapy — The relief of recognizing your life matters far less on a cosmic scale than your ego demands, freeing you to live a meaningful human-scale life.
Decide what to fail at — Strategic underachievement: pre-selecting domains where you'll deliberately not seek excellence.
Efficiency trap — The pattern where becoming more efficient surfaces more demands rather than freeing time.
FOMO / JOMO — Fear of missing out vs. joy of missing out; since missing out is guaranteed, it's what makes choices meaningful.
Helsinki Bus Station Theory — Arno Minkkinen's parable: early work resembles others' until you "stay on the bus" long enough for your route to diverge into originality.
Human disease — The compulsion to demand certainty and cosmic reassurance about the future.
Impatience spiral — The self-reinforcing loop where demanding speed erodes our tolerance for slowness, making everything feel more frustrating.
Instrumentalizing time — Treating each moment merely as a means to a future end.
Paradox of limitation — The more you chase total control over time, the worse life gets; the more you confront finitude, the better it gets.
Radical incrementalism — Robert Boice's finding: sustained creative output comes from short, fixed daily sessions, stopping on time.
Serialize — Work one big project at a time, finishing before starting the next.

Thresholds & Defaults

Lifespan budget: ~4,000 weeks (80 yrs). The number is the reframe, not a countdown to optimize.
Closed list cap: ~10 items; add only by completing.
WIP limit: ≈3 active projects (≤1 work + ≤1 non-work for big serialized projects).
Buffett split: top 5 of 25 ambitions; avoid the other 20.
Daily creative stint (radical incrementalism): small and fixed; never more than ~4 hrs; stop on time even with energy left; weekends off.
Work boundaries: set start/stop hours in advance; let the container do the deciding.
"Do nothing" practice: 5–10 min.

2026 June - What I Have Learned

Posted on 2026-06-30

Substack

Cybersecurity's AI Moment - App Economy Insights [Link]

Takeaways:

Anthropic’s Mythos model (deemed too dangerous for wide public release) has accelerated enterprise demand, forcing companies to immediately secure their AI agents and internal data systems.
Palo Alto Networks (PANW): Launched Unit 42 Frontier AI Defense and Prisma AIRS to target runtime security for AI agents. They are using internal frontier AI models to compress a year's worth of penetration testing into less than three weeks.
CrowdStrike: Created a Chief AI and Autonomous Systems Officer role (poaching from NVIDIA) and launched AIDR (AI Detection and Response), which saw ending ARR grow +250% sequentially.

Cerebras: Demand Is Not the Problem - App Economy Insights [Link]

Technical and Strategic Moats

Speed premium: Cerebras' massive "dinner-plate" Wafer-Scale Engine (CS-3) runs frontier models at over 1,000 tokens per second by avoiding the slow data transfer required between thousands of traditional GPUs.
Supply chain insulation: By utilizing mature 5nm processes and SRAM, Cerebras completely sidesteps the industry's fiercest bottlenecks—namely TSMC's CoWoS packaging and high-bandwidth memory (HBM) shortages.

Every company racing to build compute now stands in front of three doors:

Pay with the cash you already generate.

Borrow it (issue debt).

Sell a piece of the company (issue equity).

Each door sends a different signal about how confident you are and how much risk you are willing to take

― Why Google is Selling $85B of Itself - App Economy Insights [Link]

Takeaways:

Self-Funding (Free Cash Flow):
- Free cash flow is the ultimate financial position of strength and independence because it proves a company's core business can fully fund its own growth without needing anyone else's permission.
- Companies like Amazon utilize their own operating cash flow to pay for land, power, and chips directly. While this avoids interest and dilution, the massive scale of AI CapEx is pushing free cash flow down to near zero for some tech giants.
Debt Issuance:
- Debt is a highly effective leverage tool because it allows a tech giant to build much bigger infrastructure than its current cash flow allows while letting shareholders keep all the financial upside if the bet pays off.
- The Downside is it transforms a flexible bet on market demand into a rigid, inescapable financial obligation. As more AI loans move off public balance sheets into opaque special-purpose vehicles (SPVs) and private credit, the true systemic risk becomes harder to track. If AI demand drops, this extreme leverage could quickly bankrupt companies—similar to how dot-com era telecom firms collapsed after taking out 20-year loans for tech that became obsolete in just five years.
- Companies like Oracle and Meta leverage the bond market, equity-linked debt, or Special-Purpose Vehicles (SPVs) for project finance. This keeps massive liabilities off their main balance sheets and avoids immediate shareholder dilution, but it introduces substantial solvency risk if the expected AI revenue stalls.
Selling Stock (Equity):
- A company’s ownership is divided into shares, with each share representing a tiny claim on the business and its future profits. When a company issues equity, it creates brand-new slices of the pie and sells them to investors for cash.
- Unlike taking out a loan, the cash a company raises by selling stock is money it never has to pay back to lenders.
- Because the total "pie" is now cut into more pieces, existing shareholders automatically own a slightly smaller percentage of the company than they did before. This shrinking of ownership is permanent and is known as dilution, making equity the most expensive way for a company to raise capital.
- Alphabet notably chose to issue brand-new shares to raise capital while the market sits near record highs. While equity is permanently dilutive to existing shareholders, it provides non-repayable cash and signals that customer demand is outstripping what current cash reserves can build.
- The Bullish View (Market Strength): Customer and infrastructure demand is so massive that any minor shareholder dilution is irrelevant compared to the massive market opportunity and ultimate prize.
  
  The Bearish View (Risk Management): The sheer scale of the AI bet is so enormous and risky that even a cash-rich giant like Alphabet wants external partners to help share the financial downside, taking advantage of its stock sitting near an all-time high to raise cheap equity.

How FIFA Makes Money - App Economy Insights [Link]

Takeaways:

Revenue Generation (The $\$13$ Billion Machine)

IFA expects to bring in $\$13$ billion for the 2023-2026 cycle.

Core Revenue Streams (see the screenshot):
- TV Broadcasting (~40%): Remains the single largest revenue driver.
- Hospitality & Ticketing (~28%): Seeing a massive spike due to the introduction of dynamic pricing (causing top-tier tickets to climb past $\$32,000$).
- Marketing & Sponsorship (~25%): Major global brand partnerships.
Where the Money Goes (see the screenshot):

Asset-Light Execution: The 2026 World Cup relies entirely on existing stadiums (mostly NFL venues), meaning significantly fewer infrastructure costs compared to Qatar's $\$200$ billion buildout.

Redistribution Moat:
- ~58% goes to staging tournaments and events, including a record $\$871$ million prize pool (the champion takes home $\$50$ million).
- ~30% goes to development and education, which directly funds all 211 member federations. Critics view this massive financial backing as a highly durable patronage machine that secures votes for leadership.
When revenue beats expectations, the surplus flows into a reserve that is now approaching $\$2.7$ billion.
Key Risks & Controversies
- While the base ticket price remained at $60 to account for inflation, the average ticket price is estimated at $\$1,300$—a 1,000% inflation-adjusted increase since 1994, threatening the "universal" appeal of the sport
- Local governments (like Boston and Kansas City) are on the hook for tens of millions in transportation and security costs, while FIFA retains the upside from ticket sales.
- FIFA is actively trying to smooth out its volatile 4-year cycle and reduce dependency on a single tournament by introducing new properties, such as the expanded 2025 Club World Cup and targeting a $1 billion revenue mark for the 2027 Women's World Cup.

The Trillion-Dollar Off Switch - App Economy Insights [Link]

Takeaways:

Regulatory Shock: Because Anthropic could not screen users by nationality in real time, it had to disable Fable 5 and Mythos 5 worldwide. This sets a major precedent showing that a company's single most valuable frontier asset can be turned off overnight by the government.
The New Risk Variable: Choosing whether to build or rent an AI model is no longer just about cost and control; it is now about regulatory shutdown risk.
- Companies like Anthropic and OpenAI have their entire business riding on their frontier models, making them highly vulnerable to sudden regulatory shocks right as they head toward massive IPOs. Diversified companies (like Google) have other revenue streams to absorb the blow.
Apple Strategy: Instead of building a massive frontier model, Apple opted to rent a custom, 1.2-trillion-parameter version of Google’s Gemini for its new Siri AI, paying Alphabet a reported $\$1$ billion a year.
- By renting, Apple avoids the massive capital expenditure (CapEx) arms race and limits its direct exposure to model commoditization. It focuses instead on what it owns: the user's personal context and privacy.
- Apple yields control of its flagship intelligence product's roadmap and pricing to a direct competitor. Furthermore, it is only partially hedged; if Google's models get targeted by a government directive, Siri goes dark too.

Bottom Line: The core metric for AI investors is no longer just who has the most powerful model, but who can reliably keep it online within an increasingly complex regulatory perimeter.

How to Invest in IPOs - App Economy Insights [Link]

To beat the unfavorable base rates of a hot listing (like SpaceX, OpenAI, or Anthropic), follow a deliberate, patient strategy:

Let the Volatility Pass: Never rush to buy on Day 1 at peak-euphoria valuations.
Wait for the Second Earnings Call: Give the company two quarters as a public entity. This establishes a trendline, proves whether management can accurately forecast, and lets early lock-ups expire.
Size Exceptionally Small: Start with a tiny "nibble" in year one so that an early 50% drawdown won't disrupt your portfolio or peace of mind.
Anchor to Fundamentals: Base your investment on whether the valuation leaves a margin for error, not on the excitement of the narrative.

The M&A Land Grab - App Economy Insights [Link]

In the framework of the technology "land grab," companies are aggressively acquiring businesses to control specific control points before the market architecture hardens.

There are four critical layers of technology to control in a strategic race before the market structure hardens.

The Workflow Layer
- What it is: The digital environment or application where a user spends their time actually doing their primary job or task (e.g., writing code, managing a pipeline, creating a design).
- Why it matters: Sitting inside the workflow is the ultimate defensive moat because it captures the user’s active attention and creates high switching costs. If you own the workflow, you own the initial point of entry.
- The M&A Example: SpaceX’s acquisition of Cursor. Software developers live inside their IDE (Integrated Development Environment). By controlling Cursor, SpaceX/xAI captures the exact interface where developers are "vibe coding," giving them immediate influence over the tools and infrastructure developers use to build applications.
The Interface / Agent Layer
- What it is: The layer that translates user intent into execution. In the AI era, this is shifting from static buttons and menus to autonomous AI agents that can execute multi-step tasks natively on behalf of the user.
- Why it matters: The interface layer is where choices are made. If an AI agent can successfully handle a task (like resolving a customer issue or booking a flight), the user no longer needs to navigate downstream software. The agent becomes the gatekeeper to all other services.
- The M&A Example: Salesforce’s acquisition of Fin. Salesforce already owns the enterprise customer records (system of record), but by adding Fin, they gain a highly specialized, conversational AI agent layer (system of action) that can actively solve customer support issues across email, chat, and Slack without human intervention.
The Distribution Layer
- What it is: The gateway or platform that aggregates and controls audience attention, acting as the primary pipeline through which products, media, or services reach the end consumer.
- Why it matters: Without distribution, even the best content or product can be buried and rendered invisible by competing algorithms or operating systems. Whoever owns the distribution layer dictates the monetization terms, owns the first-party viewer data, and controls the ad load.
- The M&A Example: Fox’s acquisition of Roku. Fox is an expert content creator (sports, news, Tubi), but content providers are at the mercy of the hardware and operating systems people use to watch TV. By buying Roku, Fox graduates from a mere content supplier to the owner of the physical living room gateway—allowing them to control home-screen placements and monetization for over 100 million households.
The Data Loop
- What it is: A continuous feedback mechanism where product usage generates data, that data is used to improve the underlying models or product, and the improved product attracts more users.
- Why it matters: This creates a powerful flywheel effect. The company with the most active workflow or interface captures the most high-quality, real-world data, making their AI or system smarter than competitors who are relying on static or scraped datasets.
- The M&A Example: This layer underpins both tech deals (Cursor and Fin). For instance, as thousands of enterprises route customer interactions through Fin, the system maps out successful resolutions, continuously training Salesforce’s broader Agentforce platform to be more accurate and efficient over time.

Micron: Locking In the Boom - App Economy Insights [Link]

Takeaways:

Micron Technology is one of the world's largest semiconductor companies, specializing in memory and data storage technologies. Unlike tech companies that design software or processing chips (like CPUs or GPUs), Micron manufactures the physical hardware that allows computers, servers, and smartphones to hold and access data.

Revenue surged +346% year-over-year to $\$41.5$ billion, driven by skyrocketing demand for AI infrastructure. Micron’s data center business alone has cleared a $\$100$ billion annualized run rate.
Almost none of the growth came from selling more chips. Instead, due to severe capacity constraints, DRAM prices jumped over 60% and NAND prices rose in the mid-80s% in a single quarter, resulting in a staggering 85% gross margin.
To combat memory's historical cyclicality, Micron is leveraging Strategic Customer Agreements (SCAs). These are 5-year, take-or-pay contracts that lock in volumes and price floors through 2030, backed by ~$\$22$ billion in customer deposits and guarantees.
Micron’s massive margins represent a severe cost line for everyone else. Hardware giants like Apple and HP are raising device prices to absorb the memory crunch, and Micron's CEO expects supply shortages to stretch past 2027.

Narrative Violation: In B2B customer support, AI is a Copilot, Not a Replacement - a16z New Media [Link]

Takeaways:

In the business-to-business (B2B) world, AI is only resolving about 15% of tickets entirely on its own (compared to 35% in B2C, where stakes are lower). Instead of replacing humans, AI is doing the background work to make humans faster and better at their jobs.
If AI isn't closing every ticket, what is it doing? The data shows it excels at two things:
- Filtering out the noise: About 33% of all incoming support emails/messages are just "garbage" (marketing spam, automated system notifications, or accidental emails). AI quietly filters these out so humans don't waste time on them.
- Silent Triage: Two-thirds of the time, the AI reviews an incoming question, realizes it's too complex for a bot, and quietly hands it off to a human specialist with the right context attached.
When AI actively helps on a ticket—even if a human ultimately has to step in and finish it—it cuts the human's workload by about a third. Because the AI does the initial reading, sorting, and basic troubleshooting, the human support agent takes fewer messages to solve the problem than they normally would. Furthermore, customer satisfaction (CSAT) remains just as high in these hybrid human-AI setups as it does with purely human support.
The article points out that AI support gets dramatically more successful under two conditions:
1. If you give the AI access to account data (knowing who the customer is, what they bought, and their history), it resolves more tickets and gets higher ratings.
2. Companies that build their workflows around AI from the ground up have much better success rates.

Why Japanese companies do so many different things - David Oks [Link]

Takeaways:

What it is: The phenomenon of extreme corporate diversification in Japan (the "J-Firm"), where companies like Toto (toilets) or Kyocera (ceramics) successfully operate across completely unrelated, high-precision industries (such as semiconductor components).
How it works: It operates as a self-reinforcing "bundle" of horizontal practices. Instead of top-down vertical hierarchies, information flows laterally. It relies on lifetime employment, intensive generalist training, seniority-based pay, and insulation from outside shareholder pressure through insider boards and bank financing.
Why it works: Because employees cannot easily be fired and are deeply invested in the company, the primary corporate motive shifts from maximizing short-term investor profit to ensuring long-term survival. Diversification allows firms to hedge risks, reinvest earnings, and pivot to protect jobs when core markets decline.
Historical roots: The system originates from the "1940 system"—a planned, production-at-all-costs economy established by Japan during WWII mobilization. This state-controlled, employee-centric economic structure survived the war and became entrenched during the postwar era.
Strengths: Uniquely optimized for "moderate volatility" and catch-up growth. It excels at incremental, long-term shop-floor refinements, making Japan dominant in precision materials, optics, and automotive manufacturing.
Weaknesses: Highly rigid and poorly suited for sharp market discontinuities, top-down strategic shifts, and paradigm-shifting innovations (e.g., software, smartphones, and AI platforms).

train your attention span SO GOOD it feels weird to scroll - Drishti Pandita [Link]

The 5 Rules to Reset Your Attention

Don't rely on willpower. Move your phone charger to another room overnight, hide distracting apps in folders, and keep books on your bedside table.
Dedicate 30 minutes to absolute under-stimulation (no phone, no podcasts, no music). Do chores, walk, or just stare at a wall to let your baseline reset.
Do not touch your phone, email, or social media for the first hour after waking up. The circuit you fire first wins the day.
Put strict 30-minute app timers on "slot machine" feeds (Instagram, X, LinkedIn), turn off non-human notifications, and use greyscale mode.
Once a quarter, do a 7-day digital detox. Eliminate scrolling, streaming, and constant podcasts. By day 7, your baseline will recover, and you will remember what your own thoughts sound like.

What Did Claude Just Kill? 6 Moats to Protect Yourself. - Peter H. Diamandis [Link]

Anthropic is rapidly packaging Claude’s latent AI capabilities into user-friendly vertical tools—specifically targeting design, legal, and small business SaaS platforms. This "unhobbling" has triggered significant drops in valuation for traditional incumbents like Figma, Thomson Reuters, and Intuit, while fueling Anthropic's own revenue growth to a $\$30$ billion run rate.

To survive this shift, the author argues that software companies must move past thin "scaffold" applications and build defensible businesses anchored by six specific moats:

Deep Customer Relationships: Embedding deeply within organizations to make switching highly disruptive.
Proprietary Data Flywheels: Utilizing massive, unique datasets (like QuickBooks) that AI cannot easily replicate.
Trusted Brand Identity: Serving as a proxy for trust and compliance in regulated markets.
Physical-World Integration: Linking services directly to atoms, logistics, hardware, or brick-and-mortar operations.
Regulatory and Compliance Infrastructure: Navigating deeply entrenched institutional gatekeepers (e.g., FDA, HIPAA, SEC).
Network Effects: Building expansive ecosystems and platforms that become indispensable to third parties.

The Semantic Medallion: Building a Knowledge Graph-Powered Data Catalog - Veronika Heimsbakk [Link]

About transforming raw data sources into a unified knowledge graph in four lines of Python

My brutal advice to someone who wants financial freedom - Tim Denning [Link]

Takeaways:

Ditch "Sales Skills" for Credibility
- Be Authentic: Humans are hardwired to resist being sold to. Focus on building trust and being knowledgeable rather than using aggressive sales tactics.
- Walk Away: If a pitch doesn't resonate, don't force it. Move on instantly to find the right alignment.
Rational Empathy & Charisma
- "Yes, And": Use this not as a trick, but as rational empathy—reason your way to their valid position before reinforcing your own.
- Charisma: True charisma is the ability to project confidence (power) and love (good intentions) simultaneously.
- Effective Honesty: Being blunt but kind is the most effective way to communicate. If you are honest but unkind, people won't listen.
Lead, Don't Manage
- Management vs. Leadership: Management is telling people what to do; leadership is inspiring them to want to do it.
- Autonomy Breeds Freedom: Giving high-agency people a taste of freedom makes them incredibly driven, yet ultimately "unemployable" in traditional corporate structures.
Hunt in High-Trust Teams
- The "Stag Hunt" Model: Modern society thrives on high-trust cooperation. Small groups of highly competent people working together can accomplish what seems impossible. Avoid low-trust, heavily bureaucratized environments where people are forced to hunt alone.
Feed Your Obsessions
- Inspiration > Frameworks: Business books and frameworks are secondary to pure motivation. If you are genuinely obsessed and excited about a product, selling it won't feel like sales at all.
Focus on Nonlinear Upside
- Don't Fight Over Crumbs: In technology and investing, returns follow a power law (the winner takes almost everything). Don't waste time splitting hairs over a small pie; focus entirely on long-term, exponential upside.
- Preserve Your Peace: Walk away from suboptimal, highly constraining deals. Prioritize your time, reputation, and mental peace over blindly making more money.

how to enter side doors - Maja [Link]

Takeaways:

Reframe what a "job" is: A company is not just a collection of job boards and role titles; it is a group of people trying to solve a bundle of problems. A job is simply someone paying you to solve those problems.
The front door is crowded: Relying entirely on traditional job postings and resumes results in disappearing into automated, AI-saturated noise where human signal is easily lost.
Leverage two types of side doors:
- Outbound: Proactively reaching out to individuals with highly tailored, specific ideas, teardowns, or prototypes rather than a generic "can we chat?" request.
- Inbound: Creating public artifacts (essays, tools, research) so that your thinking travels ahead of you and acts as a beacon for the right people to find you.
Specificity is respect: Generic messages fail. Demonstrating that you deeply understand a person’s work, company, and precise bottlenecks creates a high-quality human signal.
Practice audacity with proof: If you want an unusual or creative career path, you must act unusually. Move forward with small, undeniable units of work without waiting for permission or a perfectly scoped job description to exist first.

Choosing to Stay Human - Ethan Mollick, One Useful Thing [Link]

The goal is not to avoid AI, but to consciously choose which tasks to automate (like rote administrative work) and which to keep human (like deep thinking and writing) to avoid losing vital mental capabilities.

Reports and Papers

Will the future of finance be shaped by talent or technology? - EY Global DNA of the CFO Survey, EY [Link]

Takeaways:

1: Why the CFO Role Should Change Now

The Challenges:
- 60% of CFOs want to define how their business creates value, but only 25% actually lead investment decisions where returns are uncertain or long-term.
- Most organizations still see finance as a control, risk, or operational function rather than a strategic partner. CFOs spend 47% of their capacity on routine operational tasks.
- 71% say traditional metrics are insufficient to evaluate initiatives that combine people and technology.
The Solutions:
- Take proactive ownership of high-uncertainty investment decisions and evolving value drivers.
- Redesign value measurement frameworks to capture the long-term impact of technology and new ways of working.
- Simplify processes, automate routine tasks, and restructure operating models to free up capacity for strategic work.

2: How Finance Can Enhance Value Creation (AI Readiness)

The Challenges:
- Only 21% of CFOs describe their function’s AI preparedness as leading or advanced.
- Securing budget for AI is heavily constrained by data quality/bias (61%), unclear or indirect long-term benefits (51%), and a lack of skills/capacity (50%).
- Most finance teams focus AI efforts on defensive applications (fraud detection, risk assessment) rather than growth-oriented applications (dynamic pricing, growth forecasting).
The Solutions:
- Prioritize underlying data quality, robust governance, and cross-functional data integration.
- Shift AI focus away from strictly defensive applications toward scalable, growth-focused, and decision-supporting capabilities.
- Embed repeatable AI capabilities directly into core finance processes.

3: Finance Transformation is About People

The Challenges:
- Only 12% of CFOs say their finance transformation outcomes exceeded expectations over the past two years; 40% reported slow or limited progress.
- Finance professionals are traditionally trained to avoid mistakes and maintain high control, which creates a cultural resistance to experimentation.
- Low Adaptability: Only 11% of CFOs describe their teams as highly adaptable, despite adaptability being a key driver of transformation success.
The Solutions:
- Invest heavily in team mindset, building a culture that safely supports change, continuous learning, and calculated experimentation.
- Build hands-on technology confidence by embedding new tools into day-to-day operations.
- Prioritize team wellbeing and manage workloads to sustain energy through transformation periods.

4: The Future CFO: New Mindsets, Skills, and Leadership Styles

The Challenges:
- While 68% of CFOs agree they must develop new leadership styles rather than rely on past expertise, 61% still rely almost exclusively on passive, self-directed learning (reading/online resources).
- CFOs rank "people and culture leadership" as only their fifth development priority, despite identifying it as their second-weakest capability.
- Underutilized Mentoring: Only 37% participate in reverse mentoring with junior staff, missing a vital avenue for learning about emerging technologies.
The Solutions:
- Prioritize firsthand leadership experience in digital transformation, complex investment decisions, and value-driver management.
- Utilize modern, interactive development pipelines such as reverse mentoring, stretch roles, and cross-functional rotations.
- Establish formal leadership pipelines and clear succession plans tied directly to ongoing transformation initiatives.

AI is creating a ‘two-track’ labour market: ‘professionalised’ roles (in which AI acts like a force multiplier for experts, requiring more human-intensive skills) see greater growth across headcount and wages than ‘democratised’ roles (in which AI makes the role itself easier for non-experts to perform)

Companies most able to use AI are seeing faster headcount growth than the least AI-exposed companies (52% vs 36%) and higher wage growth (24% vs 17%)

“Super-star companies” most exposed to AI achieved labour productivity gains of 163%, significantly outpacing other businesses

Jobs requiring specific AI skills are growing almost eight times (69%) faster than the total jobs market (9%), with the average wage premium for AI skills rising to 62%

Entry-level outlook diverges: Analysis of US data shows AI-exposed entry-level roles are seven times more likely to require traditionally senior-level skills such as judgement and leadership. These roles grew 35% since 2019, while other entry-level roles declined by 10%

― AI reshapes global labour market into two distinct paths, rewarding human skills: PwC 2026 Global AI Jobs Barometer [Link]

Two futures for jobs in an AI era - 2026 AI Global Jobs Barometer, PwC [Link]

AI for CEOs: Amplifying Time and Judgment at the Top - BCG [Link]

Core insights:

CEOs who generate meaningful value from AI spend at least eight hours a week actively building their own AI capabilities. While 72% of CEOs are directly responsible for AI decisions, only 15% are currently capturing real value.
Leaders are moving past generic, off-the-shelf tools toward customized agentic systems tailored to their specific strategic contexts, priorities, and decision-making history.
While AI improves mid-level productivity, its highest-value application is at the leadership level, expanding a CEO's most constrained resources: time and judgment.

4 Crucial Leadership Risks

Literacy $\neq$ Expertise: A polished, fluent AI response can easily mask weak assumptions, missing context, or low-confidence conclusions.
Speed $\neq$ Judgment: AI always has an answer, but compressing the path to a conclusion should not replace the human time required to absorb, challenge, and decide.
AI-Driven Groupthink: Using uniform models and identical data can reduce a leadership group's diversity of thought by up to 41%.
Cognitive Overload ("AI Brain Fry"): More automated synthesis increases the burden to review, verify, and correct data. Roughly 14% of AI users report mental fatigue from excessive oversight.

Strategic Guardrails: To ensure AI acts as a force multiplier rather than a distraction, leaders must continually evaluate:

Whether tools are genuinely exposing hidden risks and clarifying trade-offs.
If active human dissent and alternative models are built into the review process to challenge AI assumptions.
How to track and measure the long-term quality of AI-assisted decisions.

AI at Work: Strategy Matters More Than Tools - BCG [Link]

Takeaways

Around 74% of frontline employees are now regular AI users (up 23 percentage points from 2025). Among these regular users, 42% report saving an average of 8 hours per week (a full workday).
Despite massive time savings, 66% of employees receive limited or no guidance on how to spend that extra time, and more than half fail to redirect it into strategic work.
"Strategic clarity" from leadership is the ultimate differentiator. Organizations with a clear, explicit AI plan capture more value and sustain employee engagement far better than those that simply provide access to advanced tools without direction.
Awareness and integration have skyrocketed, with 30% of organizations already embedding autonomous AI agents into workflows (up from 13% in 2025). Consequently, 61% of employees believe AI agents could perform at least half of their job within the next three years.
Management structures, upskilling, and governance have not kept pace with technology. Only 36% of employees feel they have received adequate upskilling, and half of the organizations lack clear governance for managing hybrid human/AI teams.

CEO Imperatives: To succeed, leaders must move past individual tool deployment and focus on redesigning core processes end-to-end, tracking actual business value over simple adoption metrics, and personally owning the strategic vision.

From Recovery to Resurgence - BCG [Link]

Takeaways:

The global fintech industry has decisively moved past the 2023/2024 "fintech winter." Revenues grew 22% in 2025 to surpass $\$500$ billion, outpacing traditional incumbents four times over.
Growth is now defined by profitability and caution rather than 2021-style exuberance. Scaled leaders are consolidating, funding has become highly selective (Series E+ is up while seed stages contract), and public markets are imposing stricter scrutiny on IPOs.
Real near-term value is being unlocked in backend operations (engineering, fraud detection, compliance) rather than consumer-facing experiences. AI-native product development teams are delivering up to five times faster.
As consumers pivot to GenAI tools for discovery, digital marketing is shifting from Search Engine Optimization (SEO) to Generative Engine Optimization (GEO). "Recommendation-worthiness" is replacing traditional keyword bidding.
Fintechs are increasingly being pulled into banking-style regulatory perimeters globally. Navigating bank charter applications or stricter compliance models offers lower funding costs but demands higher governance standards.
While digital assets represent 15% of global fintech revenues, utility remains narrow. Stablecoins are primarily tethered to crypto trading and regional dollar access; tokenized real-world assets (money market instruments, alternatives) offer the most credible path to broader future scaling.

Articles and Blogs

What Companies Get Wrong About Decision Rights - Lindy Greer, Jennifer Jordan and Maxim Sytch, Harvard Business Review [Link]

Takeaways:

The Core Problem: Many organizations use decision-rights frameworks (like RACI, ARCI, or RAPID) incorrectly, turning them into static spreadsheets that teams ignore rather than dynamic processes that guide behavior.

The 4 Common Mistakes & How to Fix Them

Mistake 1: Confirming Roles Without Clarifying Goals
- The Issue: Defining who decides before clearly outlining what is being decided leads to ego-driven turf wars.
- The Fix: Clearly articulate specific, measurable, and time-bound subgoals first. Breaking a massive goal into smaller pieces often reveals that different stakeholders actually want to own different subgoals.
Mistake 2: Assuming Everyone Will Adhere to the Boss’s Spreadsheet
- The Issue: Mandating roles top-down or overcomplicating files with thousands of rows results in zero buy-in.
- The Fix: Co-create frameworks with the team. Use the tool as a conversation starter to air tensions and align expectations up front.
Mistake 3: Misunderstanding Roles
- The Issue: Teams frequently disagree on what roles actually mean (e.g., confusing "Accountable" with "Responsible") or invite too many people to the final decision meeting for fear of exclusion.
- The Fix: Define concrete, behavioral rules for each role:
  - Accountable (A): The single decision owner. They lead the debate and make the final call.
  - Responsible (R): A small group (2–4 people) who provide critical input and debate options. Only "A" and "R" should be in the final decision room.
  - Consulted (C) / Informed (I): Non-decision-makers who provide expertise beforehand or support execution afterward.
Mistake 4: Getting Stuck in the Same Roles
- The Issue: Executives default to holding all accountability, which burns them out and disempowers lower-level managers who possess deeper local expertise.
- The Fix: Dynamically shift roles based on the topic, not the org chart. Senior leaders should challenge themselves to hold accountability for only a few enterprise-wide decisions a year and delegate the rest.

The Power of Strategic Centering - Rita McGrath, Harvard Business Review [Link]

Takeaways:

In a "dematerializing" economy—where 90% of corporate value lies in intangibles (data, software, brands) rather than physical assets—traditional, rigid industry boundaries are collapsing. Strategic centering is the deliberate choice of one clear organizing principle to anchor a company's identity, simplify resource allocation, and enable fast, decentralized decision-making.

The 5 Strategic Centers: Companies must choose one dominant dimension to organize around:

Mission Centering: Organized around solving a massive, enduring problem, regardless of the technology used (e.g., Shopify focusing on "making commerce better for everyone" across software, logistics, and finance).
Customer Centering: Organized around deep, evolving customer needs and "jobs to be done" (e.g., Amazon starting with the customer and working backward into cloud computing, streaming, and groceries).
Technology Centering: Organized around deep, transferable capabilities to find applications across random-looking domains (e.g., Fujifilm pivoting film chemistry into cosmetics and medical imaging; Nvidia moving from gaming to AI).
Ecosystem Centering: Organized around building a critical national or regional system, often relying on public-private alignment and long-term horizons (e.g., TSMC building Taiwan’s "silicon shield").
Friction Erasure Centering: Organized around systematically making a complex domain effortless using digital tools (e.g., Toss eliminating legacy friction in South Korean financial services).

Core Benefits & Leadership Insights

A clear center allows employees to make rapid decisions and launch initiatives without waiting for corporate approval, drastically cutting bureaucratic friction and internal politics.
The most successful centered organizations are not run by hands-off managers. They require deeply engaged, intensely visible leaders (like Nvidia's Jensen Huang or Airbnb's Brian Chesky) who actively protect and reinforce the "center" daily.

6 Ways Leaders Harness Stress - Jon Miller and Drew Keller, Harvard Business Review [Link]

The 6 Stress Response Patterns

Under pressure, leaders typically default to one of six archetypes based on whether they view stress as an opportunity or a threat, and whether they react with composure or dynamic action:

The Lighthouse (Opportunity + Composed): Projects calm and stability to create psychological safety, but risks looking aloof or falling into inertia.
The Alchemist (Opportunity + Dual): Treats turbulence as fuel for innovation and growth, but can exhaust teams with constant, chaotic pivots.
The Firefighter (Opportunity + Dynamic): Thrives on adrenaline and swift execution, but can burn out teams and create impulsive, costly decisions.
The Stoic (Threat + Composed): Relies on cool logic and self-control, but risks emotionally alienating the team and internalizing stress.
The Diplomat (Threat + Dual): Uses high social intelligence to defuse conflict and unite teams, but may prioritize consensus over necessary candor.
The Container (Threat + Dynamic): Imposes rigid structure and control to filter out noise, but can isolate others by operating behind closed doors.

Key Actionable Learnings

High performance isn't about having the "perfect" default style; it’s about adaptability. Intentionally practice strategies outside your comfort zone (e.g., a Firefighter pausing to scan the horizon like a Lighthouse).
Don't assume the strategy that solved the last crisis will work on the next one. Continuously test and adjust your style based on real-time feedback.
Leaders shouldn't absorb stress alone. Build executive teams with a diverse mix of these 6 styles, and lean on a trusted circle of peers, mentors, or coaches to distribute the psychological burden.

How Agentic AI Supercharges Startups and Threatens Incumbents - Vivian S. Lee, Linda Mantia and Jon McNeill [Link]

Takeaways:

Technology is advanced to coordinated networks of specialized AI agents that can autonomously plan, act, adapt, and pursue broad organizational outcomes rather than just resolving isolated tasks.
The cost, time, and headcount needed to build, test, and pivot a startup have collapsed. Teams that used to require 6–8 people can now launch with just a domain expert and an AI engineer.
The Five Forces of Disruptive Change:
1. Digital products can be modified instantly based on feedback using natural language or "vibe-coding" tools, shrinking product-market fit timelines from over a year to mere days or weeks.
2. Startups can autonomously generate marketing content, test thousands of variants, dynamically shift ad spend, and automate complex client onboarding workflows.
3. Internal operations (HR, legal, engineering critique) can be highly automated. Internal "AI librarians" continuously index and preserve company knowledge, eliminating data fragmentation.
4. AI-native companies require up to 80% less capital and 20% to 40% less time to hit major milestones like a Series A.
5. Deep integration into customer workflows creates a self-reinforcing learning loop. More usage yields deeper data insights, improving agent performance and building high switching costs.
Established companies cannot simply bolt AI onto messy processes. They suffer from structural silos, cultural resistance from middle management, and technical debt. They must completely re-architect and simplify workflows before automating them.
AI-native startups lack the rigorous compliance, audit trails, risk mitigation, and incident-response frameworks that incumbents have spent decades refining.
The need for elite human judgment, empathy, and managing edge cases will rise. The winning future architecture is one where humans and AI agents explicitly design processes to learn together.

How Elite Sports Coaches Make High-Pressure Decisions - Alan McCall, Adrian Wolfberg, Johann Bilsborough and Ricard Pruna, Harvard Business Review [Link]

Takeaways:

Before: Disciplined Preparation Over Instinct
- Anticipate the Scenario: Elite coaches do not rely purely on gut instinct or on-the-spot reactions. They actively map out and simulate pressure scenarios before they ever occur. As one Rugby World Cup coach summarized: "You have to anticipate situations that require a decision and find the solution in advance."
- Build the Foundation Early: The quality of a high-stakes decision is often pre-determined by the level of trust, culture, and relationships established long before a crisis hits.
During: Extreme Information Filtering & Emotional Control
- Limit the Inputs: To avoid "decision paralysis" or clouding your judgment, strictly restrict incoming data to a vital few metrics or signals during crunch time. A championship-winning rugby coach highlighted a rigid cap: "Never more than three pieces of information."
- Filter Out Noise: An NBA coach similarly emphasized that receiving too much feedback from too many people leads to bad choices. Rely on focused, pre-filtered, highly valuable data streams to keep your mind clear.
- Maintain Social Awareness: Leaders must manage their own emotional control while reading the temperature of the room (or field) to maintain absolute presence.
After: Accountability and Continual Optimization
- Deconstruct the Choice: Post-crunch-time, coaches prioritize deliberate, objective reviews of the decision-making process itself rather than just looking at the outcome.
- Iterate the System: Use real-world feedback loops to update your playbook, ensuring systemic adjustments prevent the same high-pressure blind spots from repeating.

The Corporate Parallel: For business leaders, this means shifting from a reactive "firefighting" mindset to a highly structured framework where crunch-time operational variables are minimized, and crisis plays are mapped out ahead of time.

Life’s Work: An Interview with José Andrés - Alison Beard, Harvard Business Review [Link]

Takeaways:

Learn the rules like a pro so you can break them like an artist.
Flexibility beats a fixed plan. True success doesn't come from having a perfect script, but from your ability to improvise, pivot, and problem-solve the moment the unexpected happens.
No money, no mission. Passion and creativity are great, but if the bills aren't paid, the business closes. To keep doing what you love, you have to understand the numbers and run a financially tight ship.
Attitude over résumé. Hire for Passion, Not Just Skill You can train someone on technical skills, but you can't force them to care. Hire people who are naturally curious and driven, and they will figure out the rest.
Diversity and community are strengths. In business, philanthropy, and politics, the most effective way to bridge disagreements and make an impact is to invite people to sit down together, share a meal, and build empathy.

Code Isn’t Product - Richard Mironov [Link]

Takeaways:

While AI solves the problem of engineering speed, it exacerbates the problem of customer acquisition. In a market flooded with "AI slop," clear and customer-centric positioning matters more than ever.
- Clear and customer-centric positioning is defined by how you describe and frame your offering to the market. Specifically, it relies on two core principles:
  - Using the Right Language: It must be built from the exact language your actual users and prospects use to describe their perceived problems, rather than the technical language your engineers use to explain the solution.
  - Grounding in Real-World Discovery: It requires deep, systematic discovery—meaning you must actively talk to real, human customers and prospects (rather than relying on internal assumptions or "synthetic users") to understand how they experience their pain points and how they evaluate software.
The "Forward-Deployed Engineer" Symptom: The rising trend of placing dedicated engineers on-site to help enterprise customers deploy AI products is often a workaround for incomplete product design. If a customer needs a full-time engineer to realize a tool's value, the company has shipped a framework rather than a finished solution.
- The distinction:
  - A Product is self-contained. The customer can log in, understand its purpose, configure it easily, and start extracting business value with minimal hand-holding.
  - A Framework is a pile of raw building blocks. It has massive potential, but it requires a software engineer to actually build the final application before anyone can use it.
- If your product requires a resident human genius to make it work, you haven't shipped a product—you've shipped a construction site and sent along a builder.
- Mironov is warning that embedding engineers on-site is often product debt disguised as customer service. While it might keep a high-value client happy in the short term, it is a scaling bottleneck that proves the underlying software is not yet a complete, market-ready product.

Core Learnings for Product Leaders

Prioritize Real Discovery: You must talk to actual human users extensively before shipping code. Relying on synthetic users or internal assumptions will likely result in technically impressive but commercially dead-on-arrival (DOA) products.
Build Solutions, Not Toolkits: Avoid shifting the burden of product definition onto the customer or expensive, non-scalable professional services.
Double Down on Product Strategy and Marketing: As the short-term thrill of high engineering velocity wears off over the next few quarters, long-term commercial success will depend entirely on strategic product thinking at the front of the development cycle and brilliant product marketing at the back.

Self-Fulfilling Projects - Dave Hora [Link]

Takeaways:

A "self-fulfilling project" is any initiative that establishes its own importance through a highly seductive narrative rather than actual external demand, operational capacity, or objective data.

Examples:
- Korn committed to the romantic story of being a pure, self-employed craftsman who built furniture entirely by hand to maintain his creative integrity.
- "We must integrate Generative AI / Web3 / Blockchain into our core product because it is the future of our industry, and staying ahead of the curve is vital to our brand survival."
- "Our current codebase is holding us back. We need to completely pause feature development and rewrite our entire platform from scratch using a modern microservices architecture."
- "We need to restructure our entire 500-person organization into 'squads, chapters, and guilds' because this model made Spotify successful. This will make us highly innovative."
The problem: These projects feed on personal authority (often from high-level sponsors) and self-sealing logic. Because they are rooted in narrative rather than background facts, they are highly resistant to traditional, rational pushback or outside data.
The Solution: You cannot fight a self-fulfilling project from the outside. Instead, you must temporarily suspend disbelief, enter the project's internal logic, and rebuild a concrete, specific counter-structure using the language of the project itself.
Methodologies like Wardley Mapping can help turn these subjective narratives into tangible, discussable assertions that can be challenged without directly threatening a leader's authority.
- In the context of fighting "self-fulfilling projects," the biggest challenge is that you cannot defeat a compelling story with raw, outside facts. The narrative is too slippery.
  
  Methodologies like Wardley Mapping act as a solution tool because they shift the battleground. Instead of an emotional or political "story fight" between a team member and a powerful sponsor, it translates subjective narratives into a shared, visual, and challengeable map of reality.

What “done” means when you’re shipping AI features - Jeff Gothelf [Link]

Takeaways:

In traditional software development, code is deterministic. AI, however, is probabilistic. Because of this inherent unpredictability, Jeff Gothelf argues that our definition of "done" must evolve. Here is what that breakdown means in practice:

A Calibrated Distribution of Acceptable Outputs

Instead of expecting a single, binary "correct" answer (an assertion), you define success using statistical ranges (a distribution).
- The Old Way (Assertion): "When the user asks for a summary, the system returns a 3-bullet-point summary." (If it returns 4, the test fails).
- The AI Way (Distribution): "In 90% of cases, the system returns a high-quality summary. In the remaining 10%, the summary might be slightly too long or miss a minor point, but it remains coherent and safe."
Calibration is the act of actively deciding and testing what those percentages should be for your specific product and brand risk tolerance.
Managed Behavioral Variance

"Variance" is the reality that the AI will behave differently across users, prompts, and sessions. "Managed" means you aren’t just crossing your fingers and hoping for the best; you have guardrails and safety nets ready for when it inevitably acts up. This involves:
- Graceful Degradation: When the model fails, it fails safely (e.g., returning a helpful "I'm not sure" message instead of confidently hallucinating a fake stat).
- Tripwires and Monitoring: Setting up automatic alerts for key metrics (like an uptick in user thumbs-down ratings or off-tone language).
- Rehearsed Rollbacks: Having a clear, practiced protocol to instantly revert the feature or swap back to a safer model version if a tripwire is crossed.

Shipping an AI feature is no longer about proving your code is 100% perfect. It is about proving that you understand how often your AI will fluctuate, that those fluctuations are within a tolerable range, and that you have a plan ready for when it misbehaves.

Mission vs Goal: A PM’s Guide to Driving Real Impact - Aakash Gupta [Link]

Takeaways:

The article addresses a common product management trap: shipping features on time and moving isolated metrics while losing strategic coherence (the "feature factory" effect). The root cause is confusing a Mission with a Goal. Product leaders must cleanly separate the two and use an OKR chain to systematically translate broad purpose into daily engineering tasks.

The Practical Rule: If a statement requires a deadline and a dashboard, it’s a goal. If it guides trade-offs across multiple deadlines, it’s a mission.
The Translation framework: Connect the two using a 4-step framework: Mission $\rightarrow$ Themes $\rightarrow$ Objectives (1-3 directional focus areas) $\rightarrow$ Key Results (SMART metrics) $\rightarrow$ Initiatives (features/experiments).

Dimension	Mission	Goal
Core Role	Explains why the organization exists today.	Defines a specific target to pursue.
Time Horizon	Ongoing and stable.	Time-bound (e.g., quarterly).
Measurability	Qualitatively guides trade-offs; not a KPI list.	Quantifiable and measurable by design.
PM Application	Acts as a filter to say "no" to distracting work.	Represents an execution commitment.

Critical learnings for PMs:

You can successfully hit a perfectly designed SMART goal, but if it doesn't align with or advance the overarching mission, the work ultimately doesn't matter.
AI PMs often mistake model capability (e.g., "making the model better") for user value. Every feature, agent workflow, or LLM tuning layer must be tied to a metric that directly solves a real human problem.
Junior PMs talk about outputs (shipping a feature). Senior PMs talk about the outcome chain (how an initiative moved a goal that directly served the mission). Embracing this mindset is the fastest way to prove strategic maturity and clear the path for promotion.

Your AI strategy has a trust problem, not a tooling problem - Elena Verna [Link]

Takeaways:

To survive and thrive in an AI-native world, companies must shift from a model of top-down gatekeeping to one that empowers high-agency employees with the data, context, and autonomy to make rapid, distributed decisions.

Agency > Agents: AI agents are powerful tools, but they lack independent will—they wait to be told what to do. The true competitive advantage comes from high-agency employees who use AI for leverage, spot signals, and drive work forward without waiting for permission.
The Bureaucracy Tax: Heavy approval cycles, rigid title hierarchies, and bloated middle management treat employees as "risk vectors," effectively crippling the exact speed and innovation that AI tools are meant to enable.
Decisions Must Become Faster and Cheaper: Gating context forces information to travel slowly up and down the management chain. Ungating context enables daily micro-adjustments, drastically lowering the cost of making a wrong turn because teams can pivot immediately.
Middle Management is Shifting: The structural need for middle managers to act as "information telephone wires" or cross-functional referees is shrinking as information becomes centralized and democratized.

Strategic Learnings:

Just like an LLM needs data context to yield great outputs, employees need full organizational context to make correct, fast decisions.
You cannot change an entrenched corporate culture overnight with a slide deck. Instead, spin up flat, highly autonomous R&D or innovation "squads" populated by high-agency talent to prove the model works.
True employee agency requires accountability. Moving away from an "assembly-line robot" mindset means employees must be willing to own the outcomes—and the risk—of their decisions.
To build authentic trust with customers (e.g., building in public, reacting instantly to user needs), a company must first trust its own internal teams to speak and act.

How To Get Unstuck: 6 Secrets From Philosophy - Eric Barker [Link]

Takeaways:

True personal growth and breaking out of inertia do not come from waiting for the perfect mindset or overanalyzing your feelings. A better life is built from the outside in—by intentionally directing your actions, curating your attention, and managing your expectations.

Stop waiting to "feel" motivated. Act first—just put your shoes on—and the mood will eventually catch up to your momentum.
Self-esteem equals your success divided by your expectations. You don't always need to achieve more; sometimes you just need to lower the denominator by dropping ridiculous expectations.
Do something mildly annoying every day (like taking the stairs). It acts as an insurance policy for your character, building resilience for when life actually gets hard.
Wisdom is knowing exactly what not to care about. Protect your attention and stop letting digital noise squat in your brain space.
Put low-stakes decisions (like what to eat or wear) on autopilot. Save your premium mental bandwidth for the things that actually matter.
Outcomes often involve luck; effort is entirely yours. Recognizing this makes you less cruel to yourself when you fail, and less judgmental of others.

Everything is Recorded Now - a16z [Link]

Takeaways:

The integration of AI into the workplace is making the default recording of all work discussions inevitable. This shift is creating a "living context layer"—a new enterprise system of record where high-value, unstructured voice data (meetings, casual chats) is transformed by LLMs into structured, searchable insights, significantly changing how companies operate.

The best way to train an AI assistant isn't just by feeding it old wikis; it's by inviting it to meetings where it can learn company culture, expectations, and edge-case handling through "osmosis."
While privacy concerns and legal fears exist, the competitive disadvantage of not recording and leveraging this living context is massive. Controls and permissions (like "AC Priv" for sensitive meetings) will likely be retrofitted on top of widespread recording practices.

In the Age of AI, You Need a Point of View - April Dunford [Link]

Takeaways:

When breakthrough technologies like AI disrupt the market, B2B buyers become overwhelmed and hesitant to make purchases. To win their trust and business, vendors must articulate a strong, distinct Point of View (POV) about the future to reassure buyers they are making a safe, long-term investment.

A compelling POV highlights what you do better than anyone else.
If you only sell features, a confused buyer will delay their decision. You must sell your perspective on where the industry is heading.
Ensure your product choices are guided by a firm set of beliefs about the future, and communicate those beliefs clearly so the right customers can confidently choose to partner with you.

How Meta Sets Up Super IC Teams - Yue Zhao [Link]

Takeaways:

Small, cross-functional incubation teams led by highly experienced individual contributors ("Super IC teams") can drive massive business impact. However, success relies less on simply grouping top talent together and more on providing the right structural environment and operational rigor.

Executive Sponsorship: The team must report directly to an influential executive (like a CTO) who can clear roadblocks, provide real-time strategic context, and bypass normal organizational friction.
Total Self-Sufficiency: Dedicate all necessary roles entirely to the team (no part-time or borrowed resources) to maximize focus, speed, and trust.
High-Impact Focus: Do not just test random ideas. Target a large customer problem with a strict hypothesis that has a clear line of sight to significant business impact.
Pre-planned Scaling: Establish a plan and alignment for handing off the project to a larger ongoing team before the product actually needs to scale, preventing the project from stalling out.

[Fundamentals] How to share your point of view (even if you’re afraid of being wrong) - Yue Zhao [Link]

Takeaways:

Sharing your point of view is a critical way to add value at work, but many high performers hold back out of fear of being wrong.

6 Principles for Speaking Up:

Controversial ideas require a higher burden of proof.
Update your assumptions about how you actually add value.
Share where your hunch or instinct comes from.
Explain why the problem matters so people understand your motivation.
Make sure your idea makes sense on its own merit, rather than relying on your credentials.
Use language that accurately reflects your level of certainty.

Learnings:

Proximity = Insight: If you are working deeply on a problem, you might be the only person with that specific context and ability to connect the dots.
Go Beyond the Facts: Don't just document what happened or summarize data. Interpret what the facts mean for the business and share the "so what."
It's Part of the Job: You might think your role is just to pass the baton, but your part includes sharing your perspective. Your team needs to hear it.

Everyone got excited they can suddenly code, and completely missed the point - Kasper Junge [Link]

Takeaways:

While AI coding agents have made software delivery incredibly fast and cheap, they have exposed the true bottleneck in tech: deciding what to build. PMs shouldn't use AI to become one-person delivery shops. Instead, they should use it for discovery—prototyping to figure out what users actually need—because figuring out the right problem to solve is the most critical constraint in the industry.

Faster delivery doesn't fix a broken product. It just makes poor product thinking and wasteful ideas impossible to hide.
Handing developers vague tasks stripped of user context turns them into assembly-line workers and kills the team's ability to innovate.
PMs should use coding agents as discovery tools (building high-fidelity prototypes to test with users), not for shipping production features.
A PM's primary job is answering "what should we build?" Spending their time writing production code is a massive prioritization failure.
Career advice: Avoid organizations that treat software development purely as an assembly line measuring throughput. Seek out companies that focus on actual impact and innovation.

The Architecture of Focus - Magnus Hedemark [Link]

Takeaways:

Most modern work environments run on a "manager schedule" (fragmented, 30-minute blocks) that actively sabotages the deep focus required by "makers" (engineers, writers, designers). Because complex problem-solving requires long stretches of uninterrupted time just to load mental models into working memory, true productivity isn't about having more willpower—it’s about structurally redesigning your environment and calendar to make focus the path of least resistance.

Stop treating your schedule as a neutral background. Actively design it to block out sustained, uninterrupted hours for creative and technical work.
If you are struggling to focus, redesign your environment. Unplug, block apps, or go offline completely rather than relying on discipline to ignore notifications.
Keep events minimal, enforce hard deadlines, and ask yourself one simple recurring question (e.g., "What will I accomplish by Friday?") to create structural urgency.

Augmented, accelerated, autonomized: How Vanguard is embedding AI across the product lifecycle - Justin Reock [Link]

Takeaways:

Speeding up developers with AI tools doesn't guarantee faster product delivery if the rest of the team (PMs, designers, QA) is stuck at traditional speeds. The focus must shift to optimizing the entire product lifecycle to avoid an "engineering bubble."
Vanguard uses a 3-stage (Augmented, Accelerated, Autonomized) and 6-dimension AI maturity model. The ultimate goal is sweeping organizational transformation, giving teams a shared framework to evolve together.
Introducing AI agents makes basic engineering disciplines—like solid documentation, test coverage, and clean CI/CD pipelines—even more critical. At "agent speed," existing organizational dependencies and poor codebases become major bottlenecks.
AI opportunities exist well before code is written, from customer discovery to design. As AI agents take on routine execution, human roles will naturally shift toward orchestration, review, and strategic decision-making.
The biggest barrier to AI adoption is often fear and resistance to behavioral change, not the tech itself. Furthermore, simple metrics like "lines of code generated" are misleading; organizations need layered metrics that track actual business value and customer outcomes.
Proactively investing in guardrails, security, and automated governance doesn't slow things down—it actually allows teams to execute much faster and with greater confidence. Ultimately, agent speed will expose underlying organizational debt that needs fixing.

Anthropic’s Safety Superpower - Stratechery [Link]

Takeaways:

Ben Thompson argues that Anthropic's recent clash with the U.S. government over its powerful new "Fable/Mythos" model reveals the company's true driving force: a genuine, almost religious belief that they are uniquely qualified to manage AI safety. This "safety superpower" creates a perfect, but concerning, alignment where Anthropic’s aggressive business maneuvers—hoarding user data, undercutting competitors, and defying governments—are internally justified as moral necessities for protecting humanity.

Satya Nadella is actively warning against a world where a few AI models capture all economic value, pushing instead for companies to build their own proprietary "agentic systems" to retain control of their IP.

The Mom-and-Pop SaaS era has arrived - Elena Verna [Link]

Takeaways:

The article argues that the most significant impact of AI is not just increasing developer productivity, but completely democratizing software creation. As the cost and complexity of building software collapse, the barrier to entry disappears, allowing everyday professionals to build niche software solutions.

Domain Expertise > Technical Skill: The next generation of software creators will be teachers, accountants, real estate agents, and small business owners building tools for their own industries.

I built an AI that critiques me after every call. - Jenny Wanger [Link]

New Media, One Year In - a16z [Link]

Takeaways:

After its first year, a16z reflects on its New Media team, which operates as a "go-direct as a service" engine for portfolio founders. The team helps startups bypass traditional media gatekeepers to win the battle for attention. They achieve this through four pillars: high-quality in-house creative (like launch videos and essays), massive owned distribution channels (podcasts, newsletters, social media), concierge-style strategic execution, and a powerful talent network.

In today's landscrape, Founders can command attention and bypass traditional PR by being authentic, unfiltered, and relentlessly interesting.

"Insider Media" thrives because credibility is now driven by compelling characters (founders, power users, researchers) rather than faceless corporate entities.

Strategic launches and sustained media presence directly drive massive waitlists, contract value, and high-quality inbound job applications.

The Uncomfortable Reason You Keep Self-Sabotaging: 6 Secrets From Philosophy’s Most Honest Madman - Eric Barker [Link]

Takeaways:

Acknowledge the Dark Payoff: We hold onto bad habits because we secretly get an emotional kickback from them (e.g., the thrill of drama, the self-importance of being overworked).

Learning: To break a habit, you must answer honestly: What am I actually getting out of this?

Beware "Pseudoactivity": Being constantly busy is often a glamorous form of avoidance. We shuffle emails and download productivity apps to avoid doing the one terrifying thing that would actually move the needle.

Learning: Stop moving and do the scary thing first.

The "True Self" is a Myth: Waiting to find your authentic "core" causes paralysis. You are a messy combination of conflicting desires.

Learning: Stop looking inward for certainty. You build yourself outward through action, commitment, and failure.

Stop Making Leisure a Chore: Modern culture demands that we "maximize" our downtime and derive deep meaning from every weekend, turning relaxation into a performance review.

Learning: Let pleasure just be fun. It is perfectly okay if a weekend is uneventful or boring.

Ditch Toxic Hope and Resilience: Sometimes hope keeps you stuck in humiliating arrangements, and resilience just trains you to endure intolerable conditions.

Learning: Stop pretending a burning house is an "opportunity for growth." Confront brutal facts so you can actually fix them.

Reality > Fantasy: Fantasies aren't just mental vacations; they set impossible, movie-trailer standards that make your actual, messy life feel defective by comparison.

Learning: Stop comparing reality to your perfect mental movie.

Don't "Outsource" Living: We often confuse acquiring with doing. We buy the book instead of reading it, or build a massive watchlist instead of watching, tricking ourselves into feeling accomplished.

Learning: Stop collecting the accessories of a life and start the messy, three-dimensional business of actually living one.

How product management can fix your AI integration problems - Jeff Gothelf [Link]

Takeaways: Many enterprise AI pilots fail because companies drop new AI tools onto outdated workflows without changing how the actual work is done. True AI integration is a cultural and leadership challenge, not a technical one. Success requires adopting a product management approach: empowering small pilot teams to completely redesign their processes around the new technology, measuring the outcomes, and scaling what works.

YouTube

Anthropic Files For An IPO: Rapid Reaction — With M.G. Siegler - Alex Kantrowitz [Link]

Takeaways:

Public investors heavily rely on "comps" (comparable companies) to value businesses. Anthropic and OpenAI are the most direct rivals in the frontier AI space, so their financial health will be contrasted side-by-side.

Anthropic's financial narrative is currently much stronger for public markets. Anthropic is growing incredibly fast at the top line and is reported to have already dipped into profitability. OpenAI, by contrast, is burning massive amounts of cash on servers and computing power.

By jumping ahead, Anthropic effectively forces OpenAI to either rush its own public filing under intense pressure or stay private longer while its closest competitor captures the narrative—and the capital—of the public market.

Opus 4.8 Drops, Demis Hassabis Predicts AGI, and the $220B Foundation | EP #260 [Link]

Takeaways:

Google DeepMind’s Demis Hassabis has tightened his AGI timeline prediction to 2029, aligning directly with Ray Kurzweil
Amazon is winning the conversational commerce game by vertically integrating its Alexa shopping agents directly into its marketplace (boosting conversion rates 3.5x), while Google is building horizontal, open protocol layers between agents and retailers
For the first time globally, wind and solar combined have overtaken natural gas, supplying 22% of global electricity
Venture capital is aggressively pivoting from AI software toward material science, hardware, and robotics due to massive manufacturing infrastructure scales in China

OpenAI CFO Sarah Friar: IPO, AI Rivalries, New Device, and Spending $100B+ on Compute - All-In Podcast [Link]

Takeaways:

Friar emphasizes that an IPO is merely a fundraising mechanism, not the ultimate goal. OpenAI leverages multiple Cloud Service Providers (CSPs) like Microsoft Azure, Oracle, AWS, and Google Cloud to shift massive capital expenditures (capex) into operational expenses (opex), allowing them to pay for compute as they scale and generate revenue
Access to compute remains the ultimate bottleneck. Friar notes that demand is hitting a "vertical wall" and the industry will face a deficit of available tokens through 2026 and 2027
OpenAI is planning its compute needs all the way out to 2032.
OpenAI rejects the binary choice between being a consumer or enterprise brand; their revenue split is roughly 50/50
OpenAI's long-term moat is the "harness"—the layer of personal context, deep memory, and intuition built around the user (both individuals and enterprises), which makes the models highly sticky and non-commoditized
Friar teased a highly natural, seamlessly designed consumer hardware device developed with Jony Ive's team, slated for reveal by late 2026 or early 2027
While maintaining an ad-free tier, Friar noted that ChatGPT represents a potentially massive ad engine. Combining high user intent (like Google Search) with rich personal memory and context (surpassing Meta) makes it an incredibly potent environment for advertisers

Bill Ackman: Here's What the Market is MISSING - All-In Podcast [Link]

Takeaways:

Rather than focusing on short-term horizons driven by quarterly market analysts, public companies should be managed with decades or 3-to-5-year windows in mind. Ackman prefers acting as a large, stable shareholder who can support long-term initiatives even if they temporarily hurt short-term earnings.
As short-term capital rushes to chips, semiconductors, and energy, high-quality, "old-fashioned" tech giants like Microsoft, Meta, and Amazon get left behind and become undervalued. Conversely, niche SaaS companies charging high, monopolistic fees without heavy AI integration face an existential crisis

Ray Kurzweil on Why We’re Living in the Singularity | EP #261 [Link]

Takeaways:

Kurzweil reiterates his long-standing prediction that Artificial General Intelligence (AGI) will be fully achieved by 2029.
To bridge the remaining gap to AGI, Kurzweil believes AI needs progress in two specific areas: a fundamental understanding of physical world interactions (physics) and cheaper, more advanced robotics.
Most experts historically predicted AGI would take 100 years because they thought linearly. Kurzweil points out that the hardware capability has seen a 75,000 million trillion-fold increase over the last 75 years, which is why massive changes occur seemingly overnight.
The true Singularity (representing a million-fold increase in human-machine intelligence) is still on track for 2045. However, the current pace is accelerating so rapidly that systems are shifting substantially in weeks rather than years.
In the future, we won't see AI as a separate device or tool. It will seamlessly merge into our biological minds, to the point where you won't be able to distinguish whether a thought or decision originated from your biological brain or your AI enhancement.

What David Senra Learned Studying 400+ Founders - Sequoia Capital [Link]

Takeaways:

If everything is distilled into a single word, it is focus. Great founders possess an intense ability to "mute the world and build their own," completely ignoring outside consensus, critics, or what competitors are doing.
True focus isn't just ignoring bad distractions; it means saying "no" to good ideas you actually want to do because they pull you away from the great idea.
The greatest historical and modern founders (e.g., Steve Jobs, Elon Musk, Jensen Huang, Dana White) are missionaries, not mercenaries. They are fundamentally driven by the problem itself rather than purely financial returns.
You do not have to be an asshole, come from a broken home, or fit a specific psychological profile to succeed. Success comes from identifying your unique archetype (e.g., Daniel Ek of Spotify viewing himself as a "coach/team player" rather than mimicking Steve Jobs) and finding perfect founder-problem fit.
Great teams matter more than the initial idea. As Pixar's Ed Catmull noted, if you give a mediocre idea to a great team, they will either fix it or throw it out and build something completely new.
Many founders start with a dark, negative drive (a chip on their shoulder, insecurity, or a brutal inner critic). However, relying on negative self-talk forever will eventually destroy you. Long-term durability requires switching your fuel to a generative drive—building out of genuine love for the craft and a desire to create value.
True generational value compounds far into the future. Serial entrepreneurship can sometimes be overrated; selling your "life's work" too early often leaves founders unfulfilled in their later years, trying and failing to recapture that initial magic. A recurring historical maxim is simply staying in the game long enough to get lucky.
Out of more than 400 world-class founders studied, David found that only three managed to maintain what could be considered a genuinely well-balanced personal and family life. Building a historic, monolithic business almost always demands an extraordinary, lopsided sacrifice of time.

The better AI gets, the smaller its share of the economy might get – Alex Imas and Phil Trammell - Dwarkesh Patel [Link]

Takeaways:

Economists anticipate a qualitative shift where the entire supply chain of certain goods becomes fully automated, driving their human-mediated cost to zero.
As basic goods and automated tasks become virtually free, economic scarcity and value will likely shift to things where a human being in the loop is intrinsically desired (e.g., healthcare diagnostics, therapy, artistic connections, specific services)
Consumer psychology shows people are willing to pay a premium for human-vetted or human-created work, viewing AI-generated content or commodities as less valuable.

Thomas Laffont: The $\$4$T AI IPO Wave Is Coming… and We’ve Never Seen Anything Like It - All-In Podcast [Link]

Takeaways:

Funding per unicorn has increased 5x, meaning a smaller pool of companies is capturing the vast majority of capital.
A new tier of hyper-valuable private companies—including SpaceX, Stripe, Anthropic, DataBricks, Revolut, ByteDance, and Anduril—represents nearly $\$4$ trillion in value and has historically crushed the public Magnificent 7 index.
Massive, confidential S-1 public filings and upcoming listings from giants like SpaceX and Anthropic are expected to generate liquidity events that surpass the volume of the last 10 years combined.
The growth trajectories of foundational AI companies like OpenAI and Anthropic are scaling faster than any historical tech paradigm (PC, mobile, or cloud). They are on track to potentially surpass AWS and Microsoft Azure's compute scale within a few years.
The AI ecosystem is projected to hit $\$300$ billion in 2026, driven by three core pillars: consumer subscriptions, enterprise software integrations, and AI-enabled advertising (which already accounts for 25% of Meta and Google's current ad delivery).
Statistically, a private "unicorn" has only an 8% chance of becoming a $\$10$B decacorn. However, once a tech company crosses the $\$100$B "centacorn" threshold, its probability of achieving another 10x growth spurt jumps significantly to 31%, proving that winners compound exponentially.
The value of SpaceX is directly driven by its cadence of launches. As it transitions from single launches to multi-constellation platforms (like Starlink), it creates predictable, recurring revenue streams capable of capturing massive global telecom profit pools.

How We Got to the Biggest I.P.O. Race Ever | SpaceX, Anthropic & OpenAI - Hard Fork [Link]

Takeaways:

Early employees and co-founders at Anthropic and OpenAI have pledged significant portions of their equity to charity (e.g., Anthropic co-founders pledged 80%).
Over 800 mathematicians signed the Leiden Declaration out of concern that rapid AI automation will erode human norms, prioritize the "wrong" kinds of math, and flood the academic market with unchecked AI slop.
A Silicon Valley robotics startup ("The Bot Company") is facing a lawsuit for secretly renting Airbnbs to harvest domestic training data, leaving the properties heavily scuffed.
President Trump signed a voluntary AI executive order, reducing the government's model review window from 90 days down to 30 days following pushback from tech circles.
Platforms like Kalshi and Polymarket are causing "low-trust" societal issues—such as former Rep. George Santos allegedly betting against his own attendance at the State of the Union, Survivor episodes being spoiled by betting leaks, and engineers leveraging internal data.
Hackers successfully hijacked high-profile Instagram accounts (including a White House profile) simply by tricking a Meta AI support chatbot into changing account emails.

Dan Loeb: The Lost Art of Short Selling, and Why Stock Picking is Back - All-In Podcast [Link]

The IPO Comeback: Why Tech Giants Are Finally Going Public | All-In Liquidity IPO Panel - All-In Podcast [Link]

Takeaways:

Going public involves immense "garbage" and administrative overhead (e.g., massive Zoom calls, text-editing legal documents) that adds no immediate value to product engineering or sales.
While the actual day of the IPO doesn't immediately change business fundamentals, it provides significant external validation and emotional pride for long-time employees and their families.
Being a public company acts as a legitimizing event, especially when dealing with risk-averse enterprise, government, and defense clients who need proof of long-term stability.
The panel highlights a shifting trend away from the "stay private forever" philosophy. Leaders are seeing a return to companies aiming to go public earlier (at $\$1$B to $\$5$B valuations) rather than waiting until they are multi-billion or trillion-dollar monsters.
Historically, more absolute wealth is created after an IPO than before it. The panel heavily advocates for public market investors getting the chance to participate in a company's major growth stages (such as Planet Labs' 10x move in the public sphere).
While standard LLMs only understand the text of the internet, feeding them real-time geospatial satellite data will create "Large Earth Models" capable of solving real-world agricultural, climate, and security challenges.

Tony Fadell: How to build real taste (and why AI makes it matter more) - Lenny's Podcast [Link]

Takeaways:

Balancing data and opinions
1. Products require opinion, not just data: For a brand-new product category (a "1.0"), data is either non-existent or derived from old paradigms. You need a "taste-maker" or a small team to make bold, opinion-based decisions.
2. Success takes time. You make the product (Gen 1), fix the product based on user feedback (Gen 2), and then fix the business by dialing in manufacturing margins and scale (Gen 3). The iPod, iPhone, and Nest all followed this trajectory.
3. Great product ideas anchor on long-standing user pain points combined with newly emerging tech.
Marketing is part of the product
1. Technologists often get trapped explaining what a product does, but consumers care about why it matters to them. Great product building requires obsessing over how a user discovers and emotionalizes the product before they ever touch it.
2. Don't build a product and try to back-calculate the marketing later. Write the press release first to force yourself to distill the product down to its core 3 or 4 key features.
Effective micromanagement means forcing your team to get deep, precise data to build an "informed gut" for critical choices (like Apple's relentless testing to prove a software keyboard could beat BlackBerry's physical keys), rather than managing their daily workflows
AI is a powerful tool for rapid prototyping, but builders shouldn't blindly trust it to architect full systems. Doing so creates brittle, unmaintainable products and massive technical debt
Builders must design with clear principles and avoid engineering intentional user addiction or optimizing strictly for dopamine hits. True product leaders build with a focus on societal impact rather than short-term financial optimization

Why Secondary Markets Are Eating the IPO | All-In Liquidity Secondary Markets Panel - All-In Podcast [Link]

The Economics of AI Usage and What's Next For SaaS | Benedict Evans on a16z [Link]

Takeaways:

Over the past year, agentic coding has transitioned from "kind of useful" to absolute product-market fit.
AI will make building software drastically cheaper and faster, resulting in orders of magnitude more software.
A significant portion of existing SaaS companies will likely be disrupted or wiped out, which is causing widespread market hesitation and software stock derating as investors try to parse the eventual winners
The ultimate value will be determined by where the AI sits—whether it's an intelligent "bottom-of-the-stack" feature embedded seamlessly into deterministic software (like Salesforce) or a "top-of-the-stack" synthesizer across multiple company databases
Current corporate productivity gains (better analytics, faster slide creation) are difficult to quantify on a P&L statement, meaning true ROI studies will take time to settle
History shows that technology shifts start by doing old things slightly faster (e.g., more spreadsheets, quicker emails). The truly revolutionary impact occurs when businesses invent entirely new concepts that were previously cost-prohibitive or completely impossible

Learning

How to Build a Daily Web Scraper with Claude Code - Daria Cupareanu and Alex Willen, AI blew my mind [Link]

AI Agents: State, Memory, Consistency - A Deep Dive - Neo Kim and Sivasankar Natarajan, The System Design Newsletter [Link]

Neuro Linguistic Programming - Learning 2

Posted on 2026-06-23

"The Big Book of NLP, Expanded: 350+ Techniques, Patterns & Strategies of Neuro Linguistic Programming" by Shlomo Vaknin and Erickson Institute.

Outcome, Feedback, & Ecology

Ecology is NLP's term for the systemic impact of a change — checking how a desired change ripples through the whole system it lives in before you make it. The word is borrowed deliberately from biology: just as you can't change one species in an ecosystem without affecting all the others, you can't change one behavior, belief, or state in a person without affecting everything connected to it.

The skill's glossary puts it as: the state where all parts, values, and needs align with an outcome and nothing important is harmed. A change is "ecological" when it fits the whole system; it's un-ecological when it solves the stated problem but quietly breaks something else.

Before investing resources, define what you actually want — in sensory-specific, self-actionable terms — then stress-test that outcome against every part of yourself and every system you touch. An outcome no part of you objects to, and that harms no one, is the only kind that holds.

The Rule: Never commit to a change without an ecology check — surface the secondary gain and any objecting part first, or the change won't stick.

The 8 frameworks - Well-Defined Outcomes. A 10-step recipe for a "well-formed" goal answering "What do you really want?":

State it positive & specific (no "not" frames)
Frame it in terms of your own ability/actions, within your responsibility
Anchor to context — where / when / with whom, and where not
Describe in all five senses
Chunk down into achievable objectives
List the resources needed
Run an ecology check
Set observable milestones on a timeline
Write it down
Test and monitor, refining as you go

Helpful Mindsets:

Failure Into Feedback (Dilts) — break a "failure" down into what you saw, heard, and felt so it stops feeling overwhelming, mine it for the lessons it's actually teaching you, and pair it with a strong memory of success — so the setback becomes useful feedback instead of proof you can't do it.
E & E.P. Formation Pattern (Evidence & Evidence Procedure) — defines how you'll know the outcome is achieved: concrete observable evidence, the repeatable procedure to detect it (and counter-evidence), timeframes, and anticipated resistances.
Finding Positive Intention / Behavior Appreciation — a self-sabotaging habit is usually trying to do something good for you underneath. "Talk to" the part behind it, ask "What were you trying to do for me?", and keep asking until you reach the real need. Behavior Appreciation does this physically — separate floor-spots for the behavior and for the part — so you can step between them and keep them distinct.

Ecology at three scales:

Ecology Check (Grinder & Bandler) — dissociate to 3rd position; ask who benefits / who's hurt, short- vs long-term. Tool: Cartesian coordinates (do X → will/won't happen; don't do X → will/won't happen).
Secondary Gain & Personal Ecology — "What about staying stuck gives me a reason to stay stuck?"
Whole System Ecology — does this impair any other person or institution? Leverage question: "If I could have it now, would I take it?"

Mental models:

Be → Do → Have — become the person who can achieve the outcome, then act, then enjoy the result.
Failure is feedback — every miss is data + warning-knowledge (Edison's light bulbs).
A problem behavior is a messenger with a positive intention, not an enemy to suppress.

Rapport, Pacing, & Mirroring

Rapport is built by becoming subtly similar to someone — matching their body language, speech, and symbolic world — so their subconscious senses "you're like me / you get me." Once that rapport exists (pacing), you can lead them into a new state or position. Sharp sensory observation is the prerequisite for all of it.

The Rule: You can't lead someone you haven't first paced. Match their world until their subconscious accepts you as "one of us," then move — and the moment you hit resistance, that's your signal the rapport isn't there yet (back up, don't push).

The frameworks

Pacing & Matching (Grinder/Bandler) — blend elements of the other person's body language and speech into your own style (not raw imitation). Match their predicates — if they say "I see what you mean," you reply in visual language too. Start so subtle it can't be detected, escalate gradually, and never fake an accent.
Mirroring — reflect their physiology to build rapport and "second position": converse while subtly matching cadence, gesture, and breathing; use active listening ("So you're saying…") to keep it flowing; then test — try to guess their opinion before they voice it, then shift your own posture/mood and watch whether they follow (that's pacing and leading).
Behavioral & Symbolic Mirroring — when plain physical matching isn't enough (across status, gender, culture), mirror the things a person values and respects: mention hard work, family, or faith if those matter to them; match their wardrobe, references, and the connotation of their words. (The book's example: Temple Grandin wearing western wear with cattle-industry clients.)
Exchanged Matches — when copying someone directly would be too obvious, match the timing of their movement with a different, subtler one — e.g. tap your finger or nod in time with their breathing (read it from their shoulders, not their chest). Once matched, you can gently lead by slowly speeding your own rhythm so theirs follows.
When NOT to mirror — don't mirror aggression (don't play "alpha dog" — adopt a same-team posture instead), don't mirror suffering or intense need (you'll induce their state in yourself and may threaten them — mirror only the general, comforting cues). If you get caught mimicking, drop the physical mirroring but keep gentle symbolic mirroring.
The observation skills - before you can influence anyone, you have to be able to read them — and these four drills train your eyes to catch the small, automatic signals that reveal what's going on inside.:
- Eliciting Subconscious Responses — learn what each of your senses "looks like" from the outside. Ask a partner to relive a happy memory and focus first only on what they saw in it, then only what they heard, then only what they felt — while you watch how their posture, skin color, and breathing shift with each one. You're building a personal catalog of "this is what someone looks like when they're picturing something vs. hearing something vs. feeling something."
- Calibration — connect one specific body signal to one specific inner state. Have your partner think of something they clearly understand, and watch their face, eyes, and hands; then something that confuses them, and watch again. Once you can see the difference, have them silently pick one and re-think it — and you guess which from their body alone, then check. You're learning to "read the dial" for that person.
- Sensory Acuity — simply training your eyes to catch the tiny, involuntary changes people don't control: skin getting lighter or darker, muscles tensing or softening, breathing speeding up or slowing, breath moving higher or lower in the body, faint lines around the lips, pupils widening. These are the raw signals everything else is built on.
- Non-Verbal Cues Recognition — you read other people's faces better once you know your own. The face has 90+ muscles and thousands of expressions, so you practice in a mirror — cycling through emotions like fear, joy, anger, and surprise — to feel from the inside what each expression is, which makes you far quicker to spot it on someone else.
Satir Categories (Virginia Satir) — when people are under stress, they tend to fall into one of five communication styles. Each one is a kind of defense, so each needs a different approach to build rapport. Recognize which one you're facing, then match it the right way:
The Blamer — attacks and finds fault ("This is your fault"). Don't get defensive and don't fight back — instead, get worked up about the same problem they are, so you're on the same side against it. Match their intensity, but never put them in a position where they have to defend themselves.
- The Placater — over-apologizes and tries to please everyone, afraid of conflict. Give them the attention and reassurance they're craving; start by connecting with what they deeply care about, then ease into the specifics.
The Computer — goes cold, stiff, and overly logical to avoid feeling anything. Meet them there: stay calm and rational, and don't push them to open up emotionally — that only makes them retreat further.
- The Distracter — changes the subject and won't stay on point, dodging the real issue. Don't fight the tangents — go along with the detour, but keep gently steering back by tying things to what they personally care about, until their own restlessness wears them out and they return to your point.
The Leveler — straightforward, honest, and congruent: their words, tone, and body all line up. This is the easiest person to connect with — just match their honesty and sense of fairness. And when you can't tell which style someone is, default to being the Leveler yourself.

Worked example

A textile sales rep meets a clothing-company buyer. Reading her cues — worked her way up (no degree-pride), conservative-religious accent, judgmental remarks — he runs behavioral + symbolic mirroring: drops big words, mentions things he earned through hard work, uses dry humor aimed at the rich (not the poor), references his church and family. He physically mirrors her posture and breathing, but via an exchanged match — watching her shoulders and moving his hand to her breath rhythm while keeping a gentle masculine quality. Rapport forms below her awareness, and then he can lead.

State & Anchoring

A "state" is your whole mind-body condition at a given moment (confident, anxious, curious, flat). States can be deliberately switched on, cleared, amplified, and linked to a trigger — so you can re-summon a resourceful state on demand. This is the most fundamental NLP skill: nearly every other pattern assumes you can already produce and anchor a state cleanly.

Anchoring is deliberately creating a trigger that brings back a feeling on command. A song comes on and you're suddenly back in a summer ten years ago. A certain smell and you feel like a kid in your grandmother's kitchen. An anchor is a trigger (the song, the smell) got wired to a state (the feeling), so the trigger now fires the feeling automatically. NLP says, if that happens by accident, you can also do it on purpose.

The rule: Anchor at the peak, keep the state pure, and never reuse the trigger. Get that one mechanic right and most of the rest of NLP becomes available, because nearly every pattern is "elicit a state → amplify it → anchor it → fire it where you need it."

Elicit — bring up a feeling (recall a time you felt confident).
Amplify — make it stronger (turn up the mental picture, the inner voice, the body sensation).
Anchor — lock it to a trigger at its peak. (three rules: anchor at the peak (full strength), with a pure state (clean, not mixed), using a dedicated trigger (means one thing))
Fire — set off that trigger in the situation that needs it (right before the interview, the hard conversation, the stage).

The frameworks

Producing & clearing states:
- State Induction — deliberately generate a state (confidence, curiosity) before an event: define it across sight/sound/feeling, kindle it by recalling times you felt even a hint of it, amplify it through your weakest sense plus matching self-talk ("Piece of cake"), and let the feeling flow through your body.
- Accessing Resourceful States (Andreas) — for a known upcoming situation: name the quality you'll need, recall a vivid memory of having it, step in and amplify it like a "force field," then borrow a role model's physiology by viewing them from 2nd position, anchor it, and test.
- Physiomental State Interruption — snap out of a stuck bad state (boredom, anger, a self-critical voice) by taking whatever fuels it and exaggerating it until it's absurd — replay the harsh inner voice as a cartoon squeak, or jolt your mind with quick math (count back from 100 by sevens). If nothing shifts, you exaggerated the wrong detail — find the one really driving the feeling.
Anchoring — the core skill:
- Anchoring (Grinder & Bandler) — wiring a feeling to a trigger so you can call it back on command (like a song that drops you into a memory, but built on purpose): pick the feeling + a unique trigger (a hand position, a knuckle press, a private word), bring the feeling up strongly, set the trigger at its peak, then test. Keep that trigger for this feeling only — reusing it makes it meaningless.
- Self-Anchoring — the solo version: run it from the outside-observer view (watch yourself like a character in a movie, build the feeling up, then step back into your body to lock it in).
Working with stubborn negative states:
Collapsing Anchors — dissolve an unwanted automatic reaction by firing a negative-state anchor and a stronger positive-state anchor at the same time. Hold both (expect eye-darting and confusion — that's processing), release the negative first, then test. (Anchors usually go on opposite sides of the body — one per knee.)
Chaining States — when a target state is too far to reach in one jump (e.g. climbing out of a downward spiral), build a bridge of intermediate states, anchoring each to a different knuckle, then fire them in sequence from negative → positive.
Circle of Excellence (Grinder & DeLozier) — imagine a circle on the floor, load it with a peak state (step in, fully relive it, amplify, step out, test), then future-pace it: imagine stepping into it right before the real situation that needs it.
The underlying skill — noticing states: every technique above assumes you can tell what state you're in, so train that awareness first.
- Downtime — a light inward trance: turn attention inward one sense at a time (inner sounds, then images, then feelings). It's the doorway to deeper trance. (Opposite: Uptime — alert to the outside world, but still guided by inner awareness.)
- State of Consciousness Awareness — a daily journaling habit: list the states you were in, rate each, and mark what truly triggered it (often something subtle, like a tone of voice). It trains you to notice states — the foundation of every other pattern.

Submodalities & the Swish Family

Submodalities are the fine "settings" of a mental image, sound, or feeling — how bright, big, close, loud, warm it is, and where it sits. The key discovery: a feeling's intensity comes from these settings, not from the content. So you can rewire an automatic reaction by finding the few "driver" settings that hold the charge and swapping or reversing them fast. That's the engine behind the Swish and its whole family.

The rule: Find the driver submodality and turn it. A feeling is a set of adjustable dials (brightness, size, distance, location) — locate the one carrying the charge and you can shrink a craving, drain a fear, or swap a habit, all without arguing with the feeling itself.

The frameworks

The signature pattern: - catch the mental cue that sets off a bad habit and, over and over, fast, replace it with a vivid picture of the person you'd rather be — until your brain automatically goes there instead.
- The Swish (Bandler & Grinder) — replace an automatic unwanted reaction with a resourceful self-image. Build a small, dim image of the you you'd rather be; find the big, bright trigger image that sets off the bad habit; tuck the replacement as a tiny dot in its corner; then "swish" — the trigger shrinks and shoots away while the replacement explodes big and bright. Clear your mind, repeat 5–7× fast, test.
Variations on the same engine: - once you know a feeling is just a set of adjustable internal "settings," you can shrink it (Kinesthetic Swish), walk it off (Pragmagraphic), burn it out by overdoing it (Blow-Out), or spin it together with its opposite (Spinning Icons) — same engine, different gears.
Pleasure/decision tools: - Godiva borrows excitement you already have and glues it onto a task you've been avoiding; Decision Destroyer travels back to the moment a limiting decision was made and rewrites it with a strength you only gained later.
Key concepts — the handful of terms the whole chapter runs on:
- Submodality — the adjustable "settings" of a memory or mental image, sense by sense. A picture in your mind has settings like size, brightness, color, how near or far it is, where it sits, and whether it's a still photo or a moving clip. A sound has volume and pitch. A feeling has a location in your body, a temperature, and movement. These are the dials you turn to change how something affects you.
- Driver submodality — out of all those settings, the one that changes the feeling the most when you adjust it (often brightness, size, or distance). Find this one first — turning it gives you the biggest result for the least effort.
- Associated — reliving a memory from inside your own eyes, as if it's happening to you now. This carries the full emotional charge. (In the Swish, the trigger image is associated — that's why it hits hard.)
- Dissociated — watching yourself from the outside, like seeing yourself in a movie. This drains most of the emotional charge. (In the Swish, the better-self replacement is dissociated — distant enough to pull you toward it.)
- Mapping across — copying the settings from one image onto another to change how the second one feels. (E.g. take what makes a good memory feel good — bright, close, warm — and apply those same settings to a flat one.)

Neuro Linguistic Programming - Learning 1

Posted on 2026-06-12

"The Big Book of NLP, Expanded: 350+ Techniques, Patterns & Strategies of Neuro Linguistic Programming" by Shlomo Vaknin and Erickson Institute.

Vaknin frames these 21 presuppositions as "an excellent and especially useful collection."

The six that do the heaviest lifting in actual change-work: #8 (every behavior has a positive intention), #12 (meaning = the response you get), #14 (no failure, only feedback), #15 (flexibility = influence), #16 (resistance = lack of rapport), and #17 (people already have the resources).

The 21 NLP Presuppositions

The map is not the territory. — Your mental model of the world is never the world itself, and is always improvable.
People respond according to their internal maps. — To understand someone, learn their map, not what they "should" do.
Meaning operates context-dependently. — The same words/behavior mean different things in different situations.
Mind and body affect each other. — You think with your body, not just your brain; physiology and cognition are one system.
Individual skills function by developing and sequencing rep systems. — Ability is built from how you order Visual/Auditory/Kinesthetic representations.
We respect each person's model of the world. — You needn't agree, but respecting their map creates understanding and less conflict.
Person and behavior describe different phenomena — we are more than our behavior. — A single act or pattern doesn't define the whole person.
Every behavior has utility and usefulness — in some context. — Even troubling behavior carries a hidden value (basis for utilization & positive intention).
We evaluate behavior and change in terms of context and ecology. — Consider the systemic/ripple impact of any change (systems theory).
We cannot not communicate. — Clothes, posture, micro-expressions all signal; you're always communicating.
The way we communicate affects perception and reception. — Your delivery (sub-modalities, style) shapes how the message is received.
The meaning of communication lies in the response you get. — Regardless of intent, your communication is what the other person received.
The one who sets the frame for the communication controls the action. — Whoever defines the surrounding assumptions steers the exchange.
"There is no failure, only feedback." — A philosophy to live by: turn every "failure" into learning.
The person with the most flexibility exercises the most influence in the system. — Behavioral choice = control.
Resistance indicates the lack of rapport. — Resistance is a signal to rebuild rapport, not push harder.
People have the internal resources they need to succeed. — Your job is to direct them to those resources, not supply them.
Humans have the ability to experience one-trial learning. — A single intense experience can install a lasting change (good or bad).
All communication should increase choice. — Ethical use of NLP expands options; it never coerces or limits.
People make the best choices open to them when they act. — Given their map and resources at that moment, they chose what seemed best.
As responseable persons, we can run our own brain and control our results. — You can take charge of your own neurology and outcomes.

More Explanations for the Tricky Ones

Presupposition #11 — "The way we communicate affects perception and reception"

People don't just hear your words, they hear how you say them — and if your tone, style, and body don't match your message, they'll filter out the message and respond only to the delivery, so managing your delivery is what lets your content actually get through.

The core distinction: content vs. delivery

Every message you send has two layers:
- The content — the literal words, the information, the "digital message."
- The delivery — how you send it: your tone, pace, volume, posture, facial expression, word choice, the imagery you use, the emotional coloring. In NLP terms, your own submodalities and style.
This presupposition says the second layer isn't decoration on top of the message — it shapes whether the message gets through at all, and how it's understood. The book's phrasing: "You don't just communicate a digital message. You send out your own collection of sub-modalities that affect how you are perceived, and even others' ability to perceive you accurately."
The key claim: delivery can override content

This is the part most people underestimate. If your delivery clashes with your content, people filter out the content and just register the delivery. Vaknin's own example: "If you use conservative language to express liberal ideas or vice versa, people's filters will screen out the content of your message and just hear that you are conservative or liberal."

In other words — say something true in an aggressive tone, and people hear "aggression," not "truth." Deliver good news while looking anxious, and people catch the anxiety, not the news. The style becomes the message; the content gets lost behind it.

Even the rare analytical person who does parse your actual words "will probably feel uncomfortable with you, despite hearing what you were actually saying." So delivery affects two things:
- Perception — how you and your message are seen ("this person is X").
- Reception — whether the listener is able and willing to take the message in at all.
Why this happens: everyone runs filters

People don't receive your words neutrally — they run everything through their own filters (their map of the world, their mood, their assumptions about you). Your delivery is what hits those filters first. A mismatched or off-putting style trips the filter shut before your content ever lands. A well-matched style slides past the filter and lets the content in.

This is exactly why the book says the communicator's job is "to know about your listener's filters" and asks: "What are the keys that will open their minds to what you have to say?" Delivery is the key; content is what's behind the door.
The practical upshot

Three things follow if you take this seriously:
1. Manage your own state before you communicate. Your inner state becomes your delivery, and your delivery becomes their perception.
2. Match your style to your listener's filters, not to your own comfort — deliver on the channel they can receive.
3. When a true or good message keeps failing to land, suspect the delivery, not the content. You're probably broadcasting on a frequency they're filtering out.
The honest caveat

This isn't "style over substance" or a license to manipulate with slick delivery over empty content. The point is the reverse: good content deserves delivery that doesn't sabotage it. A true, valuable message delivered incongruently gets wasted — the delivery has to serve the content so it can actually be received. And bounded by Presupposition #19 (communication should increase choice), the goal of tuning delivery is to open the listener's mind to your real message, not to sneak past their judgment.

Presupposition #13 — "The one who sets the frame for the communication controls the action"

Don't walk into an important conversation arguing the content — decide first what the conversation is about and what assumptions govern it, because a frame always exists, and whoever sets it has already shaped which arguments can win.

What a "frame" is

A frame is the set of unspoken assumptions surrounding a conversation before any content is exchanged — what the discussion is about, why it's happening, what counts as relevant, what a "good outcome" would be, and who's in what role. Vaknin's own words from the book: the frame consists of "things like the assumptions about why the discussion is taking place, what the environment means about it, that sort of thing." And the key line: "Every communication comes in a package of presuppositions."
Why the frame, not the argument, controls the outcome

Most people fight over the content ("here are my five reasons…") while the other party has quietly already set the frame — and the frame decides which content even counts. A few examples of the same facts under different frames:
- A salary conversation framed as "Can we afford a raise this year?" vs. "What's the cost of you leaving and us re-hiring?" — identical facts, opposite gravity.
- A mistake framed as "Who's to blame?" vs. "What does this teach us?" — the second frame makes blame-seeking literally feel off-topic.
- A negotiation framed as "How do we split this pie?" vs. "How do we make the pie bigger?" — the frame predetermines whether it's adversarial or collaborative.
Whoever owns the frame has already won the argument that matters, because they've defined the terms on which all the smaller arguments get judged. The other person is now playing on a board someone else laid out.
"Sets the frame" — this is active, and it's usually unclaimed

The crucial insight is that a frame always exists — the only question is whether you set it on purpose or inherited one by default. The book gives the practical move directly:

"Before an important discussion, ask yourself what the frame will be if you do nothing. Consider how it may support or defeat your objectives. Then think about how that frame might be improved."

That's the whole discipline in three steps:
1. What's the default frame? If I walk in and say nothing about the terms, what assumptions will govern this? (Often the other side's, or a culturally inherited one.)
2. Does it serve or sink me? Whose objectives does the default frame favor?
3. What's the better frame, and how do I install it? — usually in the opening seconds, by naming what this is about.
The protective half

The presupposition cuts both ways, and the book stresses the defensive use. Presuppositions "can be constructive when they provide positive guidance. But they can cause harm, as they do when they serve to filter and bias propaganda. Instead of being a sitting duck, you can be proactive."

So #13 isn't only about steering others — it's about noticing the frame being placed on you. When a question contains a buried assumption ("Why does your team keep missing deadlines?" presupposes you keep missing them), answering the content accepts the frame. The skilled move is to surface and challenge the frame before engaging.

Presupposition #15 — "The person with the most flexibility exercises the most influence in the system"

Influence doesn't come from having the one right move or pushing it harder — it comes from having more moves than the situation can block, so that whenever a response fails you simply shift to the next one and keep steering toward your outcome.

Where this comes from: the Law of Requisite Variety

This is NLP's adaptation of a principle from cybernetics — Ashby's Law of Requisite Variety: in any system, the element with the widest range of responses controls the system. The skill's glossary phrases the NLP shorthand as "only variety can destroy variety" — to handle a complex or rigid situation, you need more available responses than the situation can throw at you.

"Flexibility" here means a concrete, countable thing: the number of distinct responses you can actually produce in a given moment. Not how clever or how right you are — how many different moves you have.
Why more options = more control (and not the reverse)

The intuition most people carry is backwards. We assume control comes from certainty, firmness, having the one right answer — digging in. This presupposition says control comes from the opposite: from not being locked into a single response.

The mechanism: whoever has only one response is predictable and stoppable — block that one move and they're stuck. Whoever has five responses, when the first fails, simply shifts to the second. They cannot be cornered, because the system can't exhaust their repertoire. Over any real interaction, the more-flexible party keeps adapting until they reach their outcome, while the rigid party stalls at the first obstacle.
A few concrete pictures:
- The thermostat controls the room not by being hotter or colder, but by being able to go either direction in response to whatever the room does.
- In a negotiation, the person who has only one acceptable outcome and one tactic is the one who gets stuck; the person who can switch between asking, listening, reframing, walking away, and coming back steers it.
- A parent facing a tantrum who has only "raise my voice" loses; the one who can switch to humor, distraction, curiosity, or silence finds the lever.
- A salesperson with one pitch fails on the prospect it doesn't fit; one with ten angles keeps adjusting until something lands.
"Behavioral choice = control" — the practical reframe

The actionable translation: when you feel stuck, you don't have an understanding problem, you have a flexibility problem. You've run out of responses. The question stops being "Why won't they change / why won't this work?" and becomes "What else could I do here that I haven't tried?"

This flips frustration into a generative question. Frustration is the felt sense of having exhausted your options; the cure is to manufacture another option, not to push the failing one harder. (Pushing harder is the opposite — it's reducing your variety to a single repeated move, which is why it so often fails.

Presupposition #17 — "People have the internal resources they need to succeed"

Assume the person already contains everything they need to be who they want to be in the stuck moment, and treat your role as helping them find and reconnect that resource — because a capability they retrieve themselves sticks, while one you hand them doesn't.

What "resource" means here

In NLP a resource is an internal state or capability — calm, confidence, focus, curiosity, decisiveness, compassion — that a person has already experienced at some point in their life. The claim isn't that someone literally already has the money, the skill certificate, or the job offer. It's that they already possess the neurological raw material for the state they'd need to get those things. Somewhere in their history they have been confident, been calm, been resourceful — even if not in the context where they're currently stuck.

So the presupposition is really: the problem is not a missing resource, it's a resource that isn't connected to the right context. The confidence exists; it just isn't showing up in the job interview / the difficult conversation / the moment of temptation.
"Direct, not supply" — the practitioner's actual job

This is the operative half, and it flips the usual helper role on its head.
- Supplying = "Here's what you should do, here's my advice, let me give you confidence, let me motivate you." This positions you as the source. It breeds dependence, it meets resistance, and it doesn't last, because a resource handed from outside isn't wired into the person's own neurology.
- Directing = guiding the person's mind through a process that retrieves their own existing resource and re-attaches it to the stuck context. You're a navigator, not a supplier. "You direct the client's mind through a process and let it do the work." Their brain makes the change; you just point it.
How this shows up mechanically in the patterns

Almost every technique in this book is a machine for relocating an existing resource, never for manufacturing one:
- Anchoring / Circle of Excellence — recall a past moment when you did have the state, fire it, and transfer it to where it's needed. The state already existed; anchoring just makes it portable.
- Change Personal History / Re-Imprinting — take resources the adult now has and carry them back to a past self who lacked them.
- Accessing Resourceful States — name the quality needed, find a memory of it, step in, borrow more via 2nd-position modeling.
- Six Step Reframe — even the problem behavior is reframed as containing a positive resource, not as a deficit to fill from outside.
Notice the pattern: the practitioner never installs confidence from scratch. They locate where the client has already felt it and re-route it.
Why hold this belief even if it's not literally "true"

You don't adopt #17 because it's empirically airtight — you adopt it because of the stance it forces on you:
- It makes you curious instead of prescriptive — you go looking for where the resource lives rather than lecturing.
- It keeps agency with the client — they leave knowing they did it, which makes the change durable and self-reinforcing.
- It dissolves the "broken person" frame — the person isn't deficient, they're momentarily disconnected from their own capability.

2026 April & May - What I Have Learned

Posted on 2026-05-30

Blogs and Articles

The Art of Asking Smarter Questions - Arnaud Chevallier, et al., Harvard Business Review [Link]

Using these five techniques helps us ask smarter questions by transforming our questioning from a reactive habit into a deliberate strategy. Instead of just asking whatever pops into our head, we can look at a situation and decide exactly what kind of thinking our team or project needs at that moment.

Investigative: What’s Known?
- Purpose: Digs deep to generate nonobvious information and clarify the core problem or opportunity.
- Example prompts: What happened? What is and isn't working? What evidence supports our proposed plan?
- The Goal: To avoid surface-level assumptions. It often utilizes successive "Why?" or "How?" questions to uncover overlooked data.
Speculative: What If?
- Purpose: Helps you consider a problem more broadly, reframe the issue, and explore creative, alternative solutions.
- Example prompts: What other scenarios might exist? Could we do this differently? What potential solutions have we not considered?
- The Goal: To overcome limiting assumptions and jump-start innovation (often using prompts like "How might we...?").
Productive: Now What?
- Purpose: Assesses the availability of talent, capabilities, time, and resources needed to execute a strategy.
- Example prompts: What is the next step? Do we have the resources to move ahead? Are we ready to decide?
- The Goal: To identify metrics, milestones, and potential capacity bottlenecks to keep plans on track.
Interpretive: So, What…?
- Purpose: Focuses on sensemaking and synthesis by pushing you to continually redefine the core issue and draw out the true implications of your data.
- Example prompts: What did we learn from this new information? How does this fit with our overarching goal?
- The Goal: To convert raw information into actionable insight and ensure everything ladders back up to the main mission.
Subjective: What’s Unsaid?
- Purpose: Deals with the human element—personal reservations, frustrations, hidden agendas, and emotional alignments that can derail a decision.
- Example prompts: How do you really feel about this decision? Are all stakeholders genuinely aligned?
- The Goal: To clear out "pluralistic ignorance" (where team members hide misgivings because no one else is speaking up) by creating a safe space for dissenting views.

Transformations That Work - Michael Mankins, Patrick Litre, Harvard Business Review [Link]

Summary:

Treating transformation as a continuous process: Instead of viewing change as a discrete program with a strict beginning and end (the traditional "unfreeze-change-refreeze" model), successful companies recognize that they must operate in a state of constant transformation and manage an evergreen backlog of issues.
Building it into the company's operating rhythm: Transformation shouldn't be isolated in a separate program-management office. It must be woven directly into the executive team's regular, weekly operating routines and plan reviews so that managing change becomes part of everyone's day job.
Explicitly managing organizational energy: Because transformations fail when they exhaust employees, successful programs carefully sequence initiatives. They ensure no single function or group is forced to change too many primary routines at the same time, and they reward milestones to sustain morale.
Using aspirations, not just targets, to stretch thinking: Relying strictly on external benchmarks often sets the bar too low and limits "the art of the possible." True transformations rely on bold ambitions and breakthrough thinking to fundamentally reshape business models.
Driving change from the middle out: Top-down mandates can be superficial, and bottom-up efforts often just trim around the edges. Midlevel executives possess the unique combination of frontline experience and strategic context needed to design and execute meaningful, lasting structural changes.
Accessing substantial external capital from the start: Transforming a business is expensive and cannot be reliably funded through internal cost-cutting alone. Successful efforts tap the capital markets early to ensure the initiative is fully funded to fuel growth.

Make Decisions with a VC Mindset - Ilya A. Strebulaev, Alex Dang, Harvard Business Review [Link]

To spur innovation within a traditional organization using the VC mindset, the article highlights that we don't need to apply these rules to routine decisions in predictable environments. Instead, we should deploy them strategically in times of high uncertainty, disruption, or when developing radically new products.

A High Comfort Level with Failure

The VC Skill: VCs accept that up to 80% of their investments will fail. Their business model relies on the understanding that home runs matter and strikeouts don't; one massive success can cover all other losses combined.
Prioritizing the Individual over the Group

The VC Skill: VCs seek out contrarian ideas, recognizing that breakthroughs often come from a single person in the room rather than a group. They actively structure environments to prevent individuals from being drowned out by group consensus. They achieve this by:
- Keeping teams small (often 3 to 5 partners) to ensure clear communication and accountability.
- Collecting independent feedback in advance (sometimes blindly) before meetings to avoid bias.
- Allowing junior team members to speak first so they aren't swayed by senior executives.
Valuing Disagreement over Consensus

The VC Skill: Research cited in the article shows that VC firms requiring unanimous approval actually perform worse. High-performing VCs intentionally lean into friction and unresolved opposition. For instance, some use a "consensus minus x" rule, while others (like Venrock) allow the lead partner to make the final investment decision unilaterally, regardless of other partners' skepticism. They also intentionally assign a "devil's advocate" or "red team" to argue against deals.
Championing Exceptions over Dogma

The VC Skill: VCs allow for workarounds and give individuals the power to keep a controversial idea alive even if the broader team rejects it. This includes tools like the "anti-veto," where a leader can unilaterally move a project forward despite a negative team vote to protect unconventional opportunities (as seen with early investments in Zoom and Kahoot).
Extreme Agility over Bureaucracy

The VC Skill: VCs move at an incredibly fast pace because they know slow tracking means losing the best deals to competitors. They set aggressive timelines (e.g., funding decisions within days or weeks) and operate with a streamlined, single-tier decision model.

Advice for the Unmotivated - Robin Abrahams, Boris Groysberg, Harvard Business Review [Link]

Based on the provided article, the authors outline a four-step process called DEAR (detachment, empathy, action, and reframing) to help individuals interrupt the cycle of numbness and recover their motivation at work:

Detachment

Before making major career moves or reacting under stress, you need distance and perspective to clear cognitive distortions.
- Reflect and break away: Review what went well at the end of the day, then completely disconnect from work using a physical ritual (like closing your laptop or signing out of email).
- Meditate: Dedicate 10 to 20 minutes twice a day to simple focus or breathing exercises to lower your stress response.
- Move your body: Use even brief stretches, walks, or exercise to replenish psychological energy and improve mood.
- Think in the third person: Address yourself by name or use third-person pronouns in your inner monologue to look at your problems more objectively.
Empathy

Disengagement often causes "depersonalization." Reconnecting with your humanity and the humanity of others helps rebuild motivation.
- Practice self-care: Acknowledge your values and treat yourself kindly with small daily rituals.
- Treat people as people: Combat numbness by making eye contact, practicing social niceties, and appreciating colleagues' work.
- Ask questions: Stay curious about your customers, bosses, and peers to gain fresh perspectives on your role.
- Look for friends: Cultivate real workplace friendships to make a frustrating job more enjoyable.
- Help others: Explaining systems to new hires, mentoring, or assisting teammates builds empowerment and has been shown to reduce personal burnout.
Action

Channel restless or frustrated energy into proactive, productive strategies.
- Tackle the little stuff: Checking small, mundane tasks off your to-do list builds a sense of momentum and small wins.
- Invest in outside activities: Hobbies, side hustles, or volunteering provide a sense of satisfaction that carries over into your primary job.
- Job craft: Redefine your current responsibilities by strategically shifting your focus to tasks that leverage your strengths or interests.
- Gamify: Turn boring tasks or meetings into mental puzzles, competitions, or games to trigger your competitive drive.
- Pretend and dress the part: Imagine how a mentor or superhero would handle a situation, and wear clothing that makes you feel confident and professional to help you step into character.
Reframing

Alter your perspective regarding who you are at work and why your job matters.
- Examine your work identity: Create an informal title for yourself (e.g., teacher, visionary, logistics expert) that reflects the unique value you naturally bring to a team.
- Look at the big picture: Focus on the why rather than the how of your tedious daily tasks by connecting them to a higher-order purpose or larger organizational goal.
- Consider how others benefit: Actively remind yourself of who your work serves—whether it is the clients you help directly or your family members who rely on your income.

Case Study: How Aggressively Should a Bank Pursue AI? - Thomas H. Davenport, George Westerman [Link]

Expert Recommendations

Noemie Ellezam-Danielo (Société Générale):
- Fix the Alignment: The tech team shouldn't dictate strategy alone. Siti should form a cross-functional executive group to align AI with core business goals.
- Segmented Approach: AI needs to adapt to varied customer behaviors across different demographics and geographies, rather than forcing a blanket 95% digital mandate.
- Look for Back-Office Wins: Focus on high-impact, lower-risk areas first, such as automating resource-intensive customer identity verifications.
Sastry Durvasula (TIAA):
- Embrace Disruption Safely: NVF needs "Michaels" to challenge the status quo, but the technology must supplement rather than entirely replace human service.
- Culture & Upskilling Overhaul: The shift requires an AI-savvy leadership culture. Because modern generative AI tools are user-friendly, the focus should be on upskilling existing staff to handle higher-value roles rather than just slashing head count.

Establishing a Solid Digital Foundation for AI-Everywhere Webinar - Harvard Business Review [Link]

Takeaways:

Organizations across all industries are experiencing an unprecedented wave of excitement surrounding generative AI. However, there is a stark gap between a company’s AI aspirations and its actual digital capabilities. While technology and business leaders are being pressured to deploy GenAI solutions rapidly, many lack the infrastructure to support them sustainably.
Alex Clemente reviews data from a recent HBR-AS research report (sponsored by Kyndryl) focused on the current realities of GenAI adoption.
- Many early-stage GenAI initiatives fail to scale successfully or move out of the proof-of-concept phase because organizations treat AI as an isolated tool rather than an integrated capability.
- Data quality, governance, and organizational cultural resistance remain the top barriers to meaningful adoption.
To pivot from "AI hype" to "AI success," Howard Miller and Firas Bouz highlight the strategic best practices required to build an enduring digital framework:
- Data Readiness & Architecture: Before deploying AI models, organizations must clean, unify, and secure their data architecture. AI is only as good as the underlying data feeding it.
- Modernized Infrastructure: Legacy tech stacks cannot keep up with the compute-heavy, dynamic processing needs of GenAI. Moving toward flexible, cloud-enabled or hybrid infrastructure is essential.
- Culture and Training: Overcoming user reluctance and bridging the skills gap is critical. True ROI comes from training employees to trust and integrate these technologies into their daily workflows.

Preparing for a Future Powered by Generative AI - Harvard Business Review [Link]

Transforming Consulting Through Generative AI- Harvard Business Review [Link]

Selected takeaways:

Consulting firms are seeing massive speed and efficiency benefits by positioning AI as a collaborative partner rather than a simple chatbot:
- KPMG: Has deployed gen AI capabilities to over 45,000 employees. Using the technology has cut coding assistance by 30 minutes and competitive intelligence studies by 90 minutes. Surveys show that 60% of users spend more time on high-value work, and 58% report increased creativity.
- IBM Consulting: Built a platform featuring hundreds of trained AI assistants. A task like mapping a consumer persona—which historically took a human consultant half a day—can now be generated as a baseline within 60 seconds.
- Omni Business Intelligence Solutions: Used gen AI to craft personalized proposals and marketing material, which freed up enough executive time to drive a 80% increase in closed deals year-over-year.
Succeeding with AI relies heavily on closing user intuition gaps and training knowledge workers broadly. KPMG consumer trust data details a strong generational split in technical literacy and sentiment:
- 74% of Gen Z and Millennials consider themselves highly knowledgeable about gen AI, compared to only 42% of Gen X and 13% of Boomers/Silent generations.
- To counteract this discrepancy, companies like KPMG have developed "persona-based learning journeys" led by peer "digital navigators", resulting in over 74,000 completed AI courses across their U.S. workforce.
- Similarly, PwC has invested $1 billion in gen AI, training over 75,000 employees on responsible prompt design and AI ethics to reduce anxiety surrounding job replacement.

How to Marry Process Management and AI - Thomas H. Davenport and Thomas C. Redman, Business Review [Link]

Selected takeaways:

AI typically supports narrow, specific tasks or subprocesses rather than complete, end-to-end workflows. To transform a whole operation, companies must string multiple AI use cases together. This requires heavy change management, breaking down departmental silos, and establishing common data standards across the organization.
Companies looking to merge process management with AI should follow a structured approach:

Step 1: Establish ownership. Appoint a dedicated "process owner" who can influence cross-functional teams and coordinate departments (e.g., sales, finance, operations) toward an end-to-end goal.

Step 2: Identify process customers. Determine exactly who benefits from the process (internal or external). Use generative AI to analyze customer feedback from calls, emails, and social media to find gaps.

Step 3: Map out the existing process. Use process mining and task mining software to pull real-time data from system logs. This highlights actual bottlenecks and hidden inefficiencies instead of relying on manual guesswork.

Step 4: Establish performance measures. Define clear metrics (such as cycle times or error rates) and set realistic targets based on what the data shows is possible.

Step 5: Consider process enablers. Match the right technology to the problem. Use Robotic Process Automation (RPA) for routine, repetitive tasks; use generative AI for unstructured tasks (like drafting contracts); and use traditional machine learning for predictive tasks (like fraud detection or pricing).

Step 6: Redesign the process. Utilize AI-driven design tools (like digital twins or generative AI process templates) to quickly simulate, build, and optimize new workflows.

Step 7: Implement and monitor. Continuously monitor the new process using process mining to eliminate variations, adjust to new business requirements, and maintain predictable control.

The Secret to Successful AI-Driven Process Redesign - H. James Wilson and Paul R. Daugherty, Harvard Business Review [Link]

Takeaways:

Generative AI and natural-language interfaces have made powerful tools accessible to nontechnical employees. Business transformation is no longer a niche technical skill; it is now in the hands of frontline workers who can use everyday language to surface data-rich insights and streamline their workflows.
Rather than displacing the workforce, AI acts as a tool that amplifies human creativity, experience, and intuition. Human oversight remains the critical linchpin for ensuring alignment with objectives and refining system design.
The future of continuous improvement lies in autonomous AI agents that exhibit goal-oriented behavior, logical reasoning, planning, and long-term reflection. These agents can independently make decisions, execute complex multistep workflows, and continuously adapt their strategies.
The future of continuous improvement lies in autonomous AI agents that exhibit goal-oriented behavior, logical reasoning, planning, and long-term reflection. These agents can independently make decisions, execute complex multistep workflows, and continuously adapt their strategies.
While Robotic Process Automation (RPA) is easily tripped up by process variations, ecosystems of collaborative, multimodal AI agents can handle highly complex, knowledge-intensive, and manual workflows (such as hospital revenue cycles or enterprise invoice processing) with high accuracy by learning from human demonstrations.

What People Still Get Wrong about Negotiations - Max H. Bazerman, Harvard Business Review [Link]

Takeaways:

Most negotiators mistakenly believe that a negotiation is a zero-sum game—meaning any gain for one side is an automatic loss for the other. This mindset causes people to focus strictly on claiming value rather than creating it, leaving significant profitability and mutual benefits on the table.
To avoid leaving value behind, you must identify all potential issues (beyond just the price or main percentage) and determine their relative importance before the meeting. Creating a weighted score sheet—allocating dollar or point values to different trade-offs—allows you to evaluate complex package offers logically rather than relying on emotional, on-the-spot intuition.
Resolving issues sequentially (whether starting with the easiest or hardest) is a trap. Finalizing one item before discussing the rest prevents you from discovering smart trades across different issues. Keep all items open until you can jointly map out a package that reaches the Pareto-efficient frontier (where neither side can improve without harming the other).
When trust is low or counterparts hide their cards, use these tools to uncover relative priorities:
- Build trust and share information: Cultivate an open, problem-solving dialogue (especially critical for internal corporate negotiations).
- Ask diagnostic questions: Instead of asking what they want, ask how they value issue A relative to issue B.
- Give away information first: Leverage the norm of reciprocity. Sharing your own priorities (not your walk-away bottom line) prompts the other party to share theirs.
- Make Multiple Offers Simultaneously (MESOs): Propose 3 packages of equal value to you but with different structures. The counterpart's preference will reveal their hidden priorities without them having to tell you directly.
The negotiation doesn't have to end when the deal is signed. Once an initial agreement is locked in, the adversarial tension usually drops. You can propose looking at the deal one more time to see if any new terms can be added to make it even better for both sides, with the strict ground rule that the original agreement stands unless a mutually superior alternative is found.

Leaders shouldn't Try to Do It All - A.G. Lafley and Roger L. Martin, Harvard Business Review [Link]

Takeaways:

Shift from Absolute Importance to Comparative Advantage
- The Trap: Most leaders prioritize tasks based solely on how important they are to the organization, taking on the top items until they run out of time.
- The Fix: Leaders should only spend time on activities that nobody else in the organization can do nearly as well. If someone else can do it at parity or better, it should be offloaded—even if it is a critical task.
The Four-Step Process to Manage Time

The authors suggest a rigorous framework to restructure your calendar:
- Remove absolute disadvantages: Identify "should do" tasks that you handle simply because of tradition or precedent. If a subordinate possesses equal or better skills for that specific task (e.g., routine faculty hiring or standard investor relations), hand it over completely.
- Delegate minor comparative advantages: Even if you are highly skilled at something (like creative brand reviews or routine financial management), delegate it if your advantage is only modest. This empowers your team and frees up massive blocks of your time.
- Take on strong comparative advantages: Reinvest your recovered hours into areas where you can make a unique, decisive difference. For example, A.G. Lafley focused heavily on driving a consumer-centric innovation process, while Roger Martin used his background to write books and articles that elevated his school's external profile.
- Dedicate time to the irreplaceable: Protect your calendar for high-leverage tasks that literally only the leader can execute, such as establishing massive external strategic partnerships, setting long-term brand visions, and individually mentoring the next generation of leadership.
An organization can expand its capacity, but a leader's time is fundamentally finite. Leading to win means accepting that you cannot do it all, and relentlessly focusing your energy where your personal value-add is greatest.

Want Your Company to Get Better at Experimentation? Learn Fast by Democratizing Testing - Iavor Bojinov, David Holtz, Ramesh Johari, Sven Schmit and Martin Tingley, Harvard Business Review [Link]

Takeaways:

The Necessity of Scaling Up
- High Failure Rates: Most new ideas do not yield positive results, and it is incredibly difficult to predict which ones will succeed.
- The AI Imperative: As generative AI drastically lowers the cost of creating new digital user experiences, companies must vastly increase their testing volume—into hundreds or thousands of experiments per year—to remain competitive.
Democratizing the Testing Process
- The Bottleneck: Relying solely on data scientists to design, run, and analyze every test severely limits an organization’s capacity to scale.
- The Self-Service Model: Companies must empower product managers, engineers, designers, and marketers to run their own experiments by building or buying platforms with automated guardrails, simple interfaces, and embedded statistical rigor.
- Redefining the Data Scientist's Role: Data scientists should shift from execution to high-impact work: building platforms, training teams, developing new statistical methodologies, and uncovering overarching strategic patterns.
Transitioning to "Experimentation Programs"
- Look Beyond Individual Memos: Reviewing tests in isolation wastes time and fails to build institutional knowledge.
- Learn Across Experiments: Organizations should group tests into programs (e.g., focusing entirely on "search functionality" or "product-detail pages") to identify macro-trends, catch diminishing returns, and determine where to reallocate corporate resources.
- Build a Knowledge Repository: Implement a centralized, searchable system—potentially enhanced by a generative AI assistant—to store hypotheses, metrics, and long-term insights across the entire enterprise.
Shifting Cultural and Financial Incentives
- Reward Outcomes, Not Outputs: Evaluating employees on the number of "successful" experiments makes them risk-averse. Instead, incentives should be tied to the overall performance of the business unit.
- Mitigate Risk Progressively: To encourage high-risk, high-reward testing without fear of breaking systems, companies should leverage automated rollbacks (trip wires based on guardrail metrics) and roll out features in gradual phases.

How Project Leaders Can Tame Unpredictability - Anton Skornyakov, Harvard Business Review [Link]

Takeaways:

Project leaders can tame unpredictability by using the agile technique of vertical slicing—breaking a project down into small, fully functional "slices" to run rapid tests, accelerate learning, and mitigate risks before scaling.

The author outlines how to apply this strategy across four key areas of unpredictability:

Human Behavior
- The Challenge: People's reactions to organizational change or new initiatives are highly unpredictable.
- The Takeaway: Deploy a narrowly defined, representative test group to observe real-world behavior and uncover practical issues (e.g., logistical bottlenecks, material defects) before committing a massive budget.
Interpersonal Dynamics
- The Challenge: Changes affecting career paths, organizational structures, shared responsibilities, or network effects introduce complex emotional and political dynamics.
- The Takeaway: Unlike behavioral tests, your slice here needs a large enough sample size to trigger and observe those group dynamics, but focused on a small, manageable behavioral shift.
Technological Change and Interoperability
- The Challenge: Integrating new technology with legacy systems or non-standardized external platforms is a major source of uncertainty.
- The Takeaway: Create a "tracer bullet"—a minimum viable version of the end-to-end tech solution. Manually integrate just one stream of data (e.g., a single vendor or route) all the way to the consumer-facing frontend to catch integration friction early.
Organizational Interdependencies
- The Challenge: Relying on multiple internal teams and external stakeholders introduces risks outside of your direct control.
- The Takeaway: Use an "organizational tracer bullet" to test a single operational scenario across all involved teams. This forces collaboration early and exposes contract or process hurdles before finalizing large-scale agreements.

Sam Altman May Control Our Future—Can He Be Trusted? - The New Yorker [Link]

Substack

Snowflake: AI Consumption Wins - App Economy Insights [Link]

Takeaways:

The enterprise software market is splitting between consumption-based models (which thrive as AI scales) and traditional seat-based SaaS models (which face headwinds as AI potentially replaces human roles).

Snowflake surged 38% after hours, completely shifting the narrative around its position in the AI race. Salesforce stock continued to slide as strong AI adoption metrics failed to translate into clear top-line acceleration.

The differences between Snowflake and Salesforce

Core Product & Purpose
- Salesforce is primarily a Customer Relationship Management (CRM) application layer. It is built to store and manage customer data, sales pipelines, and support workflows through an interface designed for human workers.
- Snowflake is an enterprise Data Cloud and Analytics Platform. It is a backend data warehouse and data lake layer built to ingest, store, govern, and analyze massive volumes of diverse structured and unstructured data from all corporate systems (including Salesforce, ERPs, and web logs).
Business & Pricing Model
- Salesforce traditionally operates on a Seat-Based SaaS Model. Companies pay a fixed subscription price per user license.
  - The AI Challenge: As AI agents automate tasks, companies may need fewer human "seats," threatening Salesforce's core growth model unless they can pivot to charging for AI-delivered work units.
- Snowflake operates entirely on a Consumption-Based Model. Customers only pay for the exact compute time and storage volume they use.
  - The AI Advantage: Because AI models and agents require massive amounts of compute and data ingestion to learn and execute tasks, Snowflake directly captures more revenue the harder an AI works.
Data Integration and Moat
- Salesforce operates a suite of fragmented applications (Sales, Service, Marketing, Commerce, Tableau, MuleSoft, Slack) and is trying to centralize them via its Data Cloud. It relies heavily on users inputting and managing data within its application ecosystem.
- Snowflake acts as a "Clean Sheet" repository. With its newer tools like Cortex Code, it allows organizations to run AI applications and query data that securely sits outside its own databases—including data sitting directly inside Microsoft, SAP, or Salesforce applications.

How SpaceX Makes Money - App Economy Insights [Link]

While SpaceX is seeking a massive $\$1.5$ to $\$2$ trillion valuation to raise up to $\$75$ billion, the underlying segments have dramatically different economic profiles:
- Connectivity (Starlink): The absolute profit engine. It generated $\$11.4$ billion in FY25 revenue with an impressive 63% EBITDA margin ($\$7.2$ billion). It is actively underwriting the company’s massive capital expenditures.
- Space (Launch): The strategic moat. It reflects a technical monopoly, handling over 80% of global payload weight to orbit. It shows a paper operating loss ($\$0.7$ billion) only because it completely absorbed $\$3.0$ billion in Starship R&D.
- AI (xAI, Grok, X): The capital sink. Folded into SpaceX via a February 2026 merger, this segment brought in $\$3.2$ billion in revenue but suffered a $\$6.4$ billion operating loss in FY25.
SpaceX is burning cash at an unprecedented, hyperscaler pace to build out its AI and space infrastructure:
- CapEx Ramp: Full-year FY25 CapEx hit $\$20.7$ billion. In Q1 FY26 alone, CapEx reached $\$10.1$ billion (with $\$7.7$ billion swallowed by the AI segment for gigawatt-scale clusters like COLOSSUS).
- Balance Sheet Pressure: Cash reserves dropped from $\$24.7$ billion to $\$15.9$ billion in just the first 90 days of 2026.
- The IPO's Role: This historic $\$75$ billion IPO isn't for early investor liquidity; it is strictly fuel to sustain this capital-intensive runway before Starship and AI integrations fully monetize.

Warren Buffett has long been skeptical of IPOs. His view is simple: IPOs come to market when sellers choose the timing, not when buyers are likely to get the best deal.

--- How to Invest in IPOs - App Economy Insights [Link]

Instead of falling for the fear of missing out (FOMO) on Day 1, the author suggests a "boring on purpose" strategy:

Avoid the Day 1 Hype: Let the initial open-market volatility pass completely without touching the stock.
Wait for the Second Earnings Call: The initial prospectus (S-1 filing) is just a snapshot. Waiting for two public quarters allows you to see how management handles public market scrutiny, tracks key metrics, and manages guidance.
Nibble in Year One: If you must invest, start with a tiny position. A small starter position lets you follow the company closely without making the inflated IPO price your entire cost basis.
Anchor to Valuation: Don't just look at whether a company is exceptional (like SpaceX or OpenAI)—look at whether its current stock price leaves any room for execution errors.
Give it Time: Truly great businesses will compound for a decade or more. Missing a 20% move on Day 1 is not a disaster; buying at peak euphoria usually is.

The Bottom Line: A hot IPO wave is a sign of a highly optimistic market, not an automatic buy signal. The best strategy for hyped frontier companies is to add them to your watch list, let the insider selling pressure clear, and invest strictly on your own terms.

The Trillion-Dollar Off Switch - App Economy Insights [Link]

Takeaways:

On June 12, just days after launching its powerful new "Mythos-class" model Claude Fable 5, Anthropic received a US export-control directive restricting access. Because it couldn't screen users' nationalities in real time, Anthropic took both Fable 5 and Mythos 5 offline worldwide.
The government cited national security concerns over potential jailbreaks. While Anthropic claims the vulnerability is mundane, its own defensive program (Project Glasswing) had previously shown Mythos 5 was capable of finding flaws across major operating systems and browsers, highlighting the genuine "dual-use" risk of advanced AI.
The incident highlights a new regulatory risk for pure-play AI labs (like Anthropic and OpenAI) heading toward trillion-dollar IPOs. While "building" the frontier model captures the most economic upside, it introduces a single point of failure if a government pulls the plug. "Renting" a model offers a swappable, lower-risk alternative.
Apple’s recent WWDC 2026 announcement revealed that its new Siri AI is powered by renting Google's Gemini (paying a reported $\$1$ billion/year). This strategy allows Apple to skip the massive CapEx arms race and partially insulate itself from direct model-regulation shocks, though it surrenders control over the core AI brain.

Bottom Line: For AI investors, the critical question has shifted from "Who has the best model?" to "Who can keep it online?".

How FIFA Makes Money - APP Economy Insights [Link]

Takeaways:

Driven by an expansion to 48 teams and hosting in the lucrative US market, FIFA expects to bring in a record $\$13$ billion for the 2023–2026 cycle (a 72% jump from the Qatar cycle).
Broadcasting Rights: $\$5.3$ billion (~40%) — FIFA's largest revenue engine.

Hospitality & Ticketing: $\$3.6$ billion (~28%) — Heavily boosted by 2026's new dynamic pricing model.

Marketing & Sponsorship: $\$3.3$ billion (~25%) — Major brand partnerships.

Licensing: $\$0.4$ billion (~3%) — Merchandise and video games.
As a non-profit, FIFA targets a near-breakeven budget. About 58% ($\$7.6$B) goes directly into staging competitions (including an $\$871$M team prize pool), 30% ($\$3.9$B) goes to development and education grants for its 211 member federations, and 7% ($\$0.9$B) covers governance and administrative costs.
For the first time, ticket prices float based on demand. While the cheapest seats remain around $\$60$, premium seat listings have exploded past $\$32,000$, raising concerns about pricing out everyday fans from the "people's game."
To smooth out its lumpy four-year financial heartbeat, FIFA is branching into new tournaments, notably launching the expanded 32-team 2025 Club World Cup and targeting a $\$1$ billion revenue mark for the 2027 Women's World Cup in Brazil.

Wall Street's Top Stocks in Q1 - App Economy Insights [Link]

Palantir: Tokens Are the New Coal - App Economy Insights [Link]

Takeaways:

Hyper-growth driven by the AI Platform (AIP). Dropping inference costs (cheap tokens) are massively expanding agent workflows, boosting demand for Palantir's safety and governance layer.
Revenue surged +85% Y/Y to $\$1.63$B. Remaining Deal Value (RDV) nearly doubled to $\$11.8$B.
Outlook: Management raised FY26 revenue guidance to $\$7.7$B (+71% Y/Y).

Amazon: The Inference Era - App Economy Insights [Link]

Takeaways:

AWS revenue grew +28% Y/Y to $\$37.6$ billion (its fastest growth in 15 quarters), signaling that AI monetization is hitting full stride.
Amazon is positioning AWS for the "Inference Era." While training is a GPU story, running persistent, multi-step Bedrock Managed Agents requires massive compute power, driving their custom silicon business (Graviton, Trainium) to a $\$20$ billion annualized revenue run rate.
OpenAI frontier models and Codex are entering Bedrock. OpenAI has committed to spending $\$100$ billion over 8 years on AWS and anchoring its workloads on Amazon's Trainium chips.
Amazon is maintaining a $\$200$ billion CapEx plan for 2026 to build out AI data centers. This infrastructure spend collapsed free cash flow down to $\$1.2$ billion, prioritizing long-term AI dominance over short-term cash.
Retail remains highly efficient, delivering over 1 billion same-day/overnight items so far in 2026, while Advertising surged +24% Y/Y to $\$17.2$ billion, providing a high-margin cushion for their AI investments.

Google: The Anthropic Paradox - App Economy Insights [Link]

Takeaways:

Alphabet plans to invest up to $\$40$ billion in Anthropic ($\$10$ billion upfront, $\$30$ billion in milestones) and provide 5 gigawatts of compute capacity over 5 years.
Struck at a $\$350$ billion valuation, a massive discount compared to Anthropic's recent $\$1$ trillion secondary-market price.
Even though Anthropic's Claude competes with Gemini, Google is prioritizing winning the cloud infrastructure and compute layer over model exclusivity.

Tesla: $\$25$ Billion AI Pivot - App Economy Insights [Link]

Takeaways - Tesla:

Tesla raised its FY26 capital expenditure guidance to $\$25$ billion (up from $\$9$ billion in FY25) to fund AI training clusters, "Terafab," and Optimus 3. This heavy spending is expected to turn free cash flow negative for the rest of 2026.
Q1 revenue grew 16% Y/Y to $\$22.4$ billion, and gross margin hit 21% (though aided by one-time warranty and tariff benefits).
FSD subscriptions ($\$99$/month) and Supercharger usage drove a 42% surge in Services revenue. Unsupervised Robotaxi service expanded to Houston and Dallas.
Elon Musk noted that Hardware 3 lacks the memory bandwidth for future autonomy, requiring dedicated retrofit centers and complex split software paths moving forward.

Takeaways - SpaceX:

Following its merger with xAI, SpaceX entered a strategic partnership with AI coding platform Cursor. SpaceX has the option to acquire Cursor later this year for $\$60$ billion (or pay a $\$10$ billion partnership fee if they decline).
The deal aims to fix weaknesses in xAI's Grok chatbot, which Musk admitted was "behind in coding."
Cursor gets access to SpaceX’s Colossus supercomputer (200,000 NVIDIA GPUs) to solve its compute constraints, while SpaceX secures a highly popular developer app layer and elite engineering talent.

Apple Enters a New Era - App Economy Insights [Link]

Takeaways:

In September 2026, Tim Cook will step down as CEO after 15 years, transitioning to Executive Chairman. Hardware chief John Ternus will take over as CEO.
Cook's legacy focused on operational excellence, supply chain efficiency, and high-margin Services. Ternus (an engineer by trade) represents a pivot toward new product ambitions.
Ternus inherits a $\$4$ trillion company with slowing growth (3% average since 2022). He faces high executive turnover, pressure to deliver a mass-market hardware reset (e.g., smart glasses, home robotics), and a costly supply chain diversification away from China.
Apple currently relies on Google's Gemini to power Siri. To own its AI future, Apple may need to sacrifice capital efficiency and massively ramp up CapEx, which was only $\$13$ billion in 2025 (compared to rivals planning $\$100$B–$\$200$B in 2026).

Anthropic Leapfrogs OpenAI - App Economy Insights [Link]

Takeaways:

Anthropic announced an annual revenue run rate (ARR) of over $\$30$ billion, surpassing OpenAI's reported $\$24$ billion. However, this is partly due to accounting differences; Anthropic uses a gross method (including cloud provider cuts), while OpenAI uses a net method.
Anthropic launched Mythos, a highly restricted, high-premium ($\$125$/M tokens) autonomous cybersecurity model, and secured a 3.5 gigawatt compute partnership with Google and Broadcom.
Intel partnered with Tesla, SpaceX, and xAI for Elon Musk’s 100-million-square-foot Terafab project in Austin, bringing its 18A node IP and advanced packaging technology to keep the supply chain within the US.
OpenAI acquired tech talk show TBPN for hundreds of millions to manage its narrative amid a damaging New Yorker exposé alleging safety neglect and leadership manipulation. Additionally, several top executives (CFO, COO, CMO) are stepping down or shifting roles, with internal rifts over a potential 2026 IPO timeline.
Meta launched Muse Spark (internally "Avocado"), its first closed-source model developed via its $14.3 billion Scale AI partnership. The model relies heavily on Meta’s massive distribution network (3.5 billion users) to monetize AI through consumer apps.

The Great AI Rotation - App Economy Insights [Link]

Uber’s Robotaxi Endgame - App Economy Insights [Link]

Takeaways:

Uber isn’t trying to build the best self-driving car or technology stack. Instead, it is positioning itself to be the dominant marketplace and commercialization platform that robotaxis run through.
Uber has rapidly lined up 8 autonomous partnerships (including NVIDIA, Zoox, Wayve, and Motional). Its latest move is a $\$1.25$ billion investment in Rivian to deploy up to 50,000 autonomous R2 vehicles by 2030, starting in San Francisco and Miami in 2028.

The Strategic Playbook

Through its new Uber Autonomous Solutions (UAS) branch, Uber handles the operational logistics AV developers lack: matching demand, fleet management, charging, maintenance, and financing.
Pure robotaxi fleets face severe inefficiencies during low-demand periods and unreliability during spikes. Uber layers AVs into its existing network of ~10 million human drivers to absorb volatility and ensure high asset utilization.
AV partners using Uber in early markets (Austin, Atlanta) see a 30% increase in trips per vehicle per day and 25% lower ETAs compared to launching standalone apps.

The Economic Shift

Transitioning to an AV model will likely lower Uber's traditional ~30% mobility take-rate. However, Uber expects to offset this by capturing a broader stack of autonomy service fees (ranging from 10–15% for pure distribution up to 15–25% when bundling UAS fleet services).

Key Long-Term Risk

Uber’s strategy depends on a fragmented market where tech winners still need a demand aggregator. If a competitor like Tesla (with its Cybercab) or Google's Waymo successfully scales a completely vertically integrated, direct-to-consumer network, they could bypass Uber entirely.

OpenAI Picks a Lane - App Economy Insights [Link]

Takeaways:

Shuttering the standalone Sora app and canceling a $\$1$ billion Disney partnership to cut costs and focus on high-margin enterprise seats ahead of a potential IPO.
Merging fragmented products (like Atlas browser and Codex) into a single desktop super app.
Stepping back from direct in-app payment processing to act strictly as a discovery/referral layer.

Micron: Demand Goes Vertical - App Economy Insights [Link]

Takeaways:

Q2 FY26 revenue skyrocketed 196% Y/Y to $\$23.9$ billion. Next-quarter revenue guidance (~$\$33.5$ billion) exceeds any full-year revenue in the company's history prior to 2024.
Gross margins hit 74% and are guided to 81% next quarter, edging out Nvidia's current profitability levels.
To keep up, Micron raised its FY26 CapEx forecast to over $\$25$ billion and warned that FY27 spending will step up by another $\$10$+ billion, causing some investor unease.
The memory deficit has expanded to standard PC and phone manufacturers. Industry leaders project that supply tightness could persist for four to five more years.

YouTube and Podcasts

OpenAI President Greg Brockman: AI Self-Improvement, The Superapp Bet, Path To AGI, Scaling Compute - Alex Kantrowitz [Link]

Takeaways:

OpenAI is deprioritizing standalone video generation (Sora) to concentrate its limited compute resources on a single, unified Super App that integrates chat, browsing, and advanced coding tools.
Coming later this year, OpenAI plans to deploy an automated AI researcher designed to execute the end-to-end tasks of a human research scientist in silicon, accelerating autonomous model development under human oversight.
For individuals worried about job security, Brockman's core advice is to move past the "blank box" intimidation layer and intentionally develop a sense of agency.The future workforce will thrive by acting as "CEOs" who orchestrate, delegate to, and oversee fleets of specialized AI agents.

Artemis II, Jamie Dimon’s “American Dream,” Snap’s Crucible Moment | Diet TBPN [Link]

Takeaways:

SpaceX officially filed for an IPO, immediately setting it up to enter the public markets as a trillion-dollar company.
JP Morgan Chase announced its "American Dream Initiative," a massive commitment to support small businesses, homeownership, and healthcare access. The bank plans to add 3 million new small business customers and lend up to $\$80$ billion over the next 10 years.

In a rare senior outside hire, Dimon recruited Warren Buffett’s protégé and former Geico CEO, Todd Combs. Combs will head up a new $\$10$ billion Strategic Investment Group focused on onshoring industries that America has outsourced (e.g., defense tech, US semiconductors, aerospace, and energy).
Nestle confirmed that thieves in Italy pulled off a brazen heist, stealing a delivery truck packed with 12 metric tons (413,000 units) of Kit Kat bars bound for Poland. Rather than burying the potentially embarrassing security breach, Nestle leaned into the internet humor by joking that thieves took their "Have a break" slogan too literally. This prompted other major brands (like Domino’s UK and Ryanair) to join the viral meme bandwagon, turning a logistics crisis into a corporate PR victory.

What Happens When Every CEO Becomes Omnipresent? | This Week in AI [Link]

Takeaways:

Large Tabular Models (LTMs): LLMs rely on positional encoding where the order of words matters. For deterministic database tasks (like cancer diagnosis or fraud tracking), changing column orders shouldn't change the output. LTMs solve this by focusing entirely on processing raw, structured structural data without lossy human rollups.
Nick Harris notes that chip performance no longer simply doubles every 18 months. Progress now relies on building larger chip packages and networking them together. Traditional copper cables can't travel far and are creating massive, dangerously heavy, megawatt-powered datacenter racks. Lightmatter uses glass fiber optics (photonics) to shoot 1.6 terabits per second over a single fiber, letting thousands of GPUs collaborate over a distance of a kilometer as if they were a single brain. Using optics instead of copper can triple the time it takes to train foundation models, aggressively accelerating the rate of AI takeoff.
Victor Riparbelli explains that video is currently a broadcast medium—one version shipped to everyone. Synthesia is shifting toward real-time, interactive video workflows where canvas models, synthetic diagrams, and conversational avatars roleplay and adapt to users dynamically. The current bottleneck for these immersive, custom-generated experiences is compute and bandwidth. Personalizing a one-hour custom movie right now would roughly cost $\$700$ in compute tokens.
The panel fiercely debates whether small companies should "vibe code" their own systems. Jeremy shares that his 15-person team entirely vibe-coded their own custom CRM (Fetch) inside Slack instead of buying Salesforce. Countering this, Victor and Nick argue that the hidden cost of vibe coding is the "focus cost" and token spend required for code verification. If open-source agents break or delete codebases, the developer hours spent fixing them often eclipse a simple SaaS subscription.
The fundamental difference between this wave and past industrial revolutions is that society is automating cognition rather than physical work. AGI feel hidden to the average populace because most people—including many engineers—are simply not trained to manage AI or ask deep, scientific questions.

Eric Schmidt on the Robotics Race, Singularity Timeline, and Energy Shortage | 241 - Peter H. Diamandis and Eric Schmidt [Link]

Takeaways:

Many in Silicon Valley believe that human-like AI agents and recursive self-improvement will lead to a Superintelligence moment within the next 2 to 3 years.
The ultimate resource constraint for AI scaling in America is electricity, not capital or data. There is an estimated 92-gigawatt shortage of power in the U.S. through 2030 (the equivalent of roughly 60 nuclear power plants).
Modern AI data centers have become massive airflow and liquid-cooled machines stretching half a mile long. It is projected that data centers will eventually consume 10% of the entire U.S. electricity supply.
While the U.S. is heavily focused on central AGI/ASI models, China is dominating the physical AI and low-end robotic hardware landscape, driven by their existing EV supply chain, vertical integration, and aggressive work culture.
The global landscape will likely accommodate around 10 massive-scale AI companies, primarily divided between the U.S. and China. China's strategy relies heavily on open-source edge computing.
Schmidt reiterates that a minor biological, nuclear, or catastrophic event triggered by AI may be descriptively required to force global governments (like the U.S. and China) to pause their brutal competition and cooperatively establish hard guardrails. To win the race while preserving societal health, the U.S. needs to accelerate energy permitting, actively attract high-skilled immigrants and involve experts in history, ethics, and psychology to align Superintelligence with democratic values.

SpaceX IPO, Iran War Fallout, Quantum Bitcoin Hack, The Space Opportunity - All-In Podcast [Link]

Takeaways:

Chamath predicts a 99.99% chance that Tesla and SpaceX will eventually merge into a single entity (ticker "E"). This would solve governance issues, create external market validation, and capitalize on cross-disciplinary compounding (e.g., Tesla’s autonomous robotics deployed on the Moon).
Returning to the moon via the Artemis II mission highlights a new frontier. Due to low gravity and a lack of atmosphere, manufacturing materials on the Moon and using "mass drivers" (electric rails) to shoot them back to Earth will theoretically be cheaper than traditional terrestrial shipping.
If AGI becomes a reality, the durability and valuation of software companies will erode rapidly. Investors are beginning to rotate into defensive, cash-flowing "HALO" (High Asset, Low Obsolescence) businesses.
Recent computing algorithms and theory improvements have shifted the timeline for industrial-scale quantum computing from decades away to the next 5 to 7 years.

Marc Andreessen on The Future of VC: Will a16z Go Public & Why Introspection is Dangerous? - 20VC with Harry Stebbings and a16z [Link]

Takeaways:

In venture capital, learning too much from failures can be dangerous. It is easy to completely write off a category because a previous investment failed, causing you to pass on the pattern-matched winner later on (e.g., passing on internet search in the 90s or AI before 2017). Adopting the mindset that everything is your fault drains away resentment and keeps your focus productively anchored on intrinsic self-improvement. However, as a management blanket rule for already stressed-out founders, a16z avoids pushing it.
Andreessen quotes pioneer Arthur Rock, agreeing that firms would likely perform better by shredding business plans and focusing entirely on the team/resumé. Great teams lap mediocre teams executing perfect business plans.
The Andreessen Core Formula:
- IQ: Extreme intelligence is table stakes.
- Courage: The absolute determination to run straight through problems like a cartoon character leaving a hole in a brick wall.
- Drive: An intrinsic, primal ambition to build something of your own, distinct from external markers of success like net worth.
VCs shouldn't chase "diamonds in the rough" out of investor ego. True diamonds have high visibility; if a deal has been ignored by the mainstream, it usually points to a structural defect or a hyper-disagreeable founder who has alienated everyone else. Echoing Don Valentine, more companies die from indigestion than starvation. Overfunding warps discipline, while high valuations create artificial hurdle rates that scare away future investors from down rounds.
While the remote-work era from 2020 to 2023 promised to decentralize the tech industry, AI has whiplash-reversed this trendd. The tech ecosystem is now more geographically centralized in Silicon Valley than ever before in its history.
AI is the ultimate small-D democratic technology. The highest-performing models in the world are consumerized and instantly accessible to anyone with a smartphone.
The narrative that AI causes net labor displacement is fundamentally wrong and relies on zero-sum Marxist economics. Historically and moving forward, core technology boosts the marginal productivity of the individual worker, freeing them from grunt work to perform higher-value tasks. Current tech layoffs are driven by interest rate shocks and massive, undisciplined over-hiring during COVID—not AI. Many corporate entities simply use AI as a convenient "silver bullet" excuse to downsize.

How to Reorg After AI Changes Everything | Block's Owen Jennings on the a16z Show [Link]

Takeaways:

Block eliminated the standard mid-level management layer required to coordinate traditional, heavy feature teams (which typically consisted of 14+ people, including dedicated project managers, product managers, and numerous localized engineering leads). They replaced these with highly fluid squads of 1 to 6 people. Because these squads leverage automated AI tooling, they require significantly fewer coordinators and middle managers to align their work.
Product and engineering management layers were cut by 50% to 60%, widening leadership spans and drastically accelerating the internal velocity of data and decision-making.
Workflows transitioned from linear code production and manual peer reviews to asynchronous oversight. Managers and engineers run dozens of parallel background agents, context-switching between them to review, nudge, and commit agent-generated pull requests.

Why Juries Are Turning Against Meta and YouTube - Hard Fork [Link]

Takeaways:

Separate juries in Los Angeles and New Mexico recently found social media giants liable for harming young users
A recent leak exposed the agentic framework surrounding Anthropic's coding harness (Claude Code), enabling developers to rapidly clone and "Frankenstein" the tool over open-source models.
Citing research from Anthropic's Nicholas Carlini, AI tools have become so proficient at finding software vulnerabilities that nearly all legacy codebases (like the Linux kernel) will eventually need to be hardened to resist AI-driven cyber attacks.

How Bots, Deepfakes and AI Agents Are Forcing a New Internet Identity Layer | Alex Blania on a16z [Link]

Takeaways:

The exponential rise of highly persuasive AI agents and deepfakes is making it nearly impossible to digitally distinguish real humans from bots, rendering standard one-to-one biometric verification obsolete.

To eliminate dystopian privacy risks, World uses Multi-Party Computation to fragment iris codes across decentralized computers and relies on zero-knowledge proofs so users can verify their uniqueness anonymously.

Without a cryptographically strong infrastructure to verify humanity, AI-driven automation will scale massive fraud across government social programs and structurally collapse the ad-based creator economy through bot-driven content farms.

World is shifting 90% of its resources to an aggressive U.S. expansion, aiming to deploy 50,000 biometric "Orbs" in major retail spaces and pilot on-demand motorbike couriers to ensure users are within 15 minutes of a device.

OpenClaw, Claude Code, and the Future of Software | Peter Yang on The a16z Show [Link]

Takeaways:

Peter explains that 70–80% of the value of OpenClaw for him is its personal, conversational aspect. Using voice interactions via messaging apps like Telegram feels much more direct and personal than traditional chat interfaces.
Functional apps used primarily to complete specific tasks (like Mercury banking or Calendly) are seeing reduced manual usage. Users prefer instructing an agent to run analytics, manage documents, or interact with APIs on their behalf.
AI code generation tools currently feel like slot machines or casinos due to variable wait times and unexpected results. However, once customized with personal hooks and skills, they become deeply integrated workflows.
Functional apps used primarily to complete specific tasks (like Mercury banking or Calendly) are seeing reduced manual usag. Users prefer instructing an agent to run analytics, manage documents, or interact with APIs on their behalf.
AI is highly capable of tackling the first 80% of a creative or technical project (such as writing code or blog drafting), while human input is strictly needed to manually tweak and polish the final 20%.
Instead of scaling corporate headcount, future companies will prioritize staying as lean as possible. A traditional 10-person product team may shrink down to two or three humans paired with a network of specialized AI agents.
Entrusting objective negotiations and tedious operational functions (like long OKR alignment meetings) to AI agents eliminates high-emotion corporate back-and-forth.
While large corporations may continue trimming staff, the cost to build software or launch a business is dropping close to zero. This enables a rise in independent solopreneurs, bootstrapped micro-businesses, and creators who can build their own products without traditional engineering backgrounds.

The State of Modern War: Palantir & Anduril Execs on Drones, AI, and the End of Traditional Warfare - All-In Podcast [Link]

Takeaways:

America has dangerously hollowed out its commercial manufacturing ecosystem, leaving a hyper-consolidated defense industrial base that lacks the raw volume and agility required to rapidly regenerate weapon stockpiles during a major conflict.
Modern tech startups are bypassing the government's sclerotic, "cost-plus" procurement bureaucracy by using private venture capital to build advanced, merit-based products before pitching them to the state.
Integrating AI into the command center introduces unprecedented targeting precision that dramatically minimizes civilian casualties compared to legacy "dumb bombs," though human operators must always remain accountable for the system's actions.
Failing to aggressively re-industrialize the West risks yielding a "Chinese Century" where the CCP explicitly dictates the global terms of international trade and engagement.

OpenAI's TBPN Mistake, SpaceX’s $2 Trillion IPO?, Iran Disables Amazon Infrastructure - Alex Kantrowitz [Link]

Takeaways:

OpenAI acquired the daily tech talk show TBPN (The Better Podcasting Network)
SpaceX has filed for an IPO, aiming to publically raise between $40 billion and $80 billion—potentially making it three times larger than any historical IPO
In internal memos, Amazon Web Services (AWS) confirmed that Iranian-backed strikes in Bahrain and Dubai physically took down critical "availability zones,". Due to the severe volatility, the cost of specialized war and terrorism insurance for data center physical assets in the Gulf region has skyrocketed by 1,900% in just a few weeks.

Why Balaji Srinivasan Thinks the SaaS Apocalypse Is Overhyped | The a16z Show [Link]

Takeaways:

The AI economy will shift toward distillation, decentralization, and localized "trusted tribes" resembling the low-trust Chinese internet model rather than central tech monopolies.
AI text and slides are frequently lazy, generic, or deceptive ("Lauram AIPSM"). AI drastically reduces the cost of generation but scales up the hidden cost of validation.
Humans will remain the essential "sensors" of taste and agency, while AI serves strictly as the "actuator" under a leash. AI turns employees into CEOs rather than replacing them.
The wholesale destruction of SaaS is overhyped. Intelligent incumbents will survive by leveraging their existing distribution loops.

While vulnerable incumbents that fail to execute may be cloned or disrupted, dominant developer platforms like Notion or Figma can use AI to build features and ship value to their established networks significantly faster. A raw technical clone of an app like Facebook yields zero value without its underlying user graph and distribution.
Bitcoin has evolved strictly into transparent institutional collateral rather than digital cash for individuals. Zcash (and wallets like Zodal) will fulfill the actual multi-decade vision of private, scalable digital cash.

Meta Tokenmaxxing, Intel Joins Terafab, Frontier AI vs. China | Diet TBPN [Link]

We Have to Talk About Anthropic's Mythos - Hard Fork [Link]

Takeaways:

Anthropic developed an incredibly powerful new model but has held it back from the public due to extreme cybersecurity risks. Instead of a commercial launch, Anthropic initiated Project Glasswing, giving exclusive, restricted access to the "blue teams" (defensive cybersecurity units) of major tech infrastructure and platform companies—such as Cisco, Broadcom, Microsoft, Apple, and Amazon.
The hosts note that the sheer volume of undiscovered zero-day bugs this model can chain together autonomously means almost every major piece of software globally may need to be systematically patched or rewritten, creating a massive human bottleneck for software maintainers.
This marks the first time since GPT-2 in 2019 that a massive gap has opened between the frontier capabilities hidden inside top AI labs and what the general public can actually use
While the systemic threat is institutional, the hosts emphasize that everyday users should protect themselves against localized fallout by locking down basic digital hygiene: using password managers for unique, randomly generated passwords and enforcing multi-factor authentication (MFA) via authenticator apps.

The Mistake That Could Break America - David Friedberg - Chris Williamson [Link]

Takeaways:

Friedberg highlights that a massive portion of leaders in the technology sector are actively leaving or preparing to leave California.
Friedberg argues that politicians get elected by promising benefits they ultimately cannot fund. California faces a massive pension and healthcare liability crisis for public employees, estimated to be between $\$600$ billion and $\$1$ trillion in the hole.
The state's massive tax revenues have been funneled into poorly managed projects. Friedberg points to massive budget waste, such as a $\$30$ billion high-speed rail project with rampant executive turnover and a $\$220$ million homeless program that only successfully transitioned six individuals out of poverty.
The core philosophical threat discussed is California's proposed "Billionaire Tax Act". Friedberg emphasizes that taxing post-tax assets creates a dangerous precedent where private property rights are systematically degraded.
While pitched at a 5% rate for billionaires, Friedberg warns that historical precedents—like the original US income tax, which started in 1913 as a mere 1% tax on high earners —prove that thresholds inevitably drop until a tax applies to the broader middle class. He fears the end state is one where 51% of the population votes to seize everything from the other 49%.
Friedberg notes a tragic irony: humanity is right on the cusp of an era of unprecedented technological abundance—including near-free energy, radical life extension, and automated labor. Despite this potential, public sentiment remains deeply pessimistic. Friedberg cites a poll showing that AI is currently the most unfavorable concept in the United States, even ranking below highly polarizing political figures. He warns that if Western regulatory and tax regimes stifle this technological path, global competitors like China will glean the benefits instead.

What Really Happened Behind Closed Doors at OpenAI - Hard Fork [Link]

Takeaways:

The New Yorker's Profile on Sam Altman

Journalists Ronan Farrow and Andrew Marantz discussed their massive 16,000-word profile examining whether OpenAI CEO Sam Altman can be trusted.
The piece portrays Altman not through a single "smoking gun" incident, but rather a long, subtle accumulation of complaints from colleagues claiming he repeatedly tells different audiences different things.
The hosts reveal that the law firm investigation commissioned after Altman's brief 2023 firing was never actually put into writing; it was kept entirely verbal to limit liability, resulting only in a vague 800-word press release.
While the profile is quite damning, it also highlights that Altman is the target of intense, sometimes fabricated smear campaigns orchestrated by tech rivals like Elon Musk.

More OpenAI C-Suite Drama, Is Siri Seriously Broken?, Meta’s Elusive Next Hit - Alex Kantrowitz [Link]

What You Can Learn From Our $\$20$B Groq Deal - Chamath Palihapitiya [Link]

Takeaways:

Traditional corporate org charts are often "bullsh*t" built for internal politicians; real value is driven by building the company around elite technical talent
Startups require teams to find emotional reward in technical milestones and quiet progress during years of darkness before hitting exponential growth
True "A-level" performance comes from a culture of self-directed, high-agency individuals pushing each other—"iron sharpens iron".

Box CEO on the AI Adoption Gap | The a16z Show [Link]

Takeaways:

Instead of relying on predefined IT connections, agents will seamlessly orchestrate data across complex, multi-system enterprise software (like SAP or Workday) at runtime
You cannot simply "vibe code" your way through massive legacy software architectures. Systems of record aren't just neat data sets; they carry decades of complex business logic, meaning legacy systems are not disappearing anytime soon
Wall Street models are missing the exponential growth curve by viewing the market through a zero-sum lens. Just like mainframes, bandwidth, and cloud compute before it, algorithmic or hardware shifts will drastically collapse the cost of intelligence, making current token-metering anxieties temporary.

Anthropic’s $30B Ramp, Mythos Doomsday, OpenClaw Ankled, Iran War Ceasefire, Israel's Influence - All-In Podcast [Link]

Josh Shapiro on Trump, Iran War Chaos, Israel's Failure, the Economy, and 2028 Race - All-In Podcast [Link]

Marc Andreessen introspects on Death of the Browser, Pi + OpenClaw, and Why "This Time Is Different" - Latent Space and a16z [Link]

SpaceX Goes Public, Claude’s Mythos Release, and the US Data Center Delay | EP #246 - Peter H. Diamandis [Link]

Ben Horowitz on AI Anxiety, Big Tech Transitions & The Future of Startups | a16z [Link]

Elon Musk vs. Sam Altman, AI Job Loss, and OpenAI’s $852B Valuation | EP #247 [Link]

Jensen Huang – Will Nvidia’s moat persist? - Dwarkesh Patel [Link]

Everything You Know Is About to Collapse - David Friedberg - Chris Williamson [Link]

The Pentagon's AI Plan + Behind the Anthropic Fight — With Under Secretary of War Emil Michael - Alex Kantrowitz [Link]

Building Agents at Home: Homeschooling, Parenting and More | The a16z Show [Link]

Allbirds’ AI Pivot, Snap Cuts 16% of Workforce, Amazon’s GlobalStar Deal | Diet TBPN [Link]

OpenAI's Identity Crisis, Datacenter Wars, Market Up on Iran News, Mamdani's First Tax, Swalwell Out - All-In Podcast [Link]

Why the A.I. Backlash Turned Violent in America - Hard Fork [Link]

I also feel like I had a reasonably prepared mind so that when that door opened, I was able to walk through it with my own two feet.

― Why Reading Most Books Is A Waste Of Time - Chamath Palihapitiya [Link]

Why Washington Suddenly Wants A.I. Regulation - Hard Fork [Link]

Takeaways:

Anthropic's unreleased preview model, Claude Mythos, is the primary reason Washington is suddenly panicking. The model has proven to be incredibly powerful at finding novel code vulnerabilities and daisy-chaining them together.
1. Palo Alto Networks CEO Nikesh Arora noted that during a concerted auditing effort using these advanced models, his company discovered seven times the volume of exploits they would typically find in a normal period.
2. While proprietary SaaS can be patched quickly, older legacy code and open-source software remain high-risk vectors. Cyber security firms are using A.I. to build perimeter firewall "scaffolding" to block automated attacks before software can be manually updated.
3. The Pentagon is legally battling Anthropic in court—designating it a "supply chain risk" because the lab refused a contract allowing "any lawful use"—while simultaneously installing and implementing Claude Mythos to scan their own systems for security vulnerabilities.
4. The sudden consensus that "something must be done" has triggered a fierce bureaucratic scramble within Washington over who gets to hold the keys to A.I.
The hosts highlight that while Washington is scrambling toward safety regulations, its current execution is deeply fractured and contradictory:
1. At the exact same time national security officials are warning that these models are too dangerous to fall into adversary hands, Trump invited Nvidia CEO Jensen Huang onto Air Force One to join a high-profile trade delegation to China. The trip's goal includes negotiating chip trade policies—essentially trying to sell China the raw hardware compute power needed to build the very caliber of models (like Mythos) that Washington is trying to restrict.
2. This institutional confusion is perfectly mirrored in the Pentagon, which has been using Claude Mythos to scan and secure its defensive infrastructure while simultaneously fighting Anthropic in court to label the company a "supply chain risk" because the lab refuses to allow unrestricted military use of its commercial tech.
Venmo is moving away from its famous public-by-default feed, shifting onboarding for new users to "friends only".
Frugal Amazon employees are allegedly abusing the company's internal A.I. tool (Meshclaw) to generate unnecessary background activity simply to boost their token consumption and look more productive to leadership.
A University of Central Florida commencement speaker was loudly booed by humanities and arts graduates after proclaiming A.I. as the "next industrial revolution".

How the 1% Will Own Compute (and What It Means for You) - This Week in AI and This Week in Startups [Link]

Anthropic CEO on Safety, Job Displacement and Anthropic's $350B Valuation | WSJ [Link]

Takeaways:

Amodei emphasizes that while public sentiment fluctuates wildly between AI hype and a burst bubble every 3–6 months, the actual capability of AI models follows a smooth, exponential line upward—acting essentially as a "Moore's Law for intelligence".
The signature of this AI transition will likely be a world experiencing very high GDP growth (5% to 10%) occurring simultaneously with high unemployment (potentially 10%) and increased inequality. Anthropic launched the Anthropic Economic Index to track in real-time exactly how Claude is automating versus augmenting tasks across industries so that policymakers aren't acting blindly.
Anthropic intentionally prioritizes business clients over consumer spaces to avoid the toxic incentives of traditional tech, like maximizing user engagement, serving ads, or generating algorithmic "slop". Developers and non-technical staff alike have shifted heavily toward agentic workflows rather than simple text chat, allowing tools to build entirely new apps end-to-end or organize projects within minutes. Anthropic is explicitly choosing not to focus on photo or video generation, viewing much of consumer short-form video AI content as addictive "slop" with minimal value to enterprise operations.
One of Amodei's major worries is a dystopian outcome where a highly contained "zeroth world country" of ~10 million tech elites (mostly centered in Silicon Valley) accelerates to 50% GDP growth while decoupling entirely from the rest of the slower-moving global economy.
When asked about the missing piece to make frontier AI safe, Amodei pointed to mechanistic interpretability—the science of physically looking inside neural networks (similar to an MRI for the human brain) to confirm if a model is being deceptive or lying, rather than just relying on text testing.

Watts, Wafers, and the Future of AI Infra | Gavin Baker - Invest Like The Best [Link]

Google I/O 2026 keynote in 35 minutes - The Verge [Link]

Joe Rogan Experience #2501 - Marc Andreessen - PowerfulJRE [Link]

Is It Time to Break the Two-Party System? | The Ezra Klein Show [Link]

Why MKBHD is glad he never went viral | EP 53 - Hard Fork [Link]

Claude Code Head Boris Cherny: Insane Growth, Tokenmaxxing, AI Agents' Next Frontier - Alex Kantrowitz [Link]

Sundar Pichai on Whether Google Is Falling Behind in A.I. - Hard Fork [Link]

SpaceX’s $2T Case, Nvidia’s Shock Selloff, America Turns on AI, Trump Pulls AI Order, Bond Crisis? - All-In Podcast [Link]

2026 March - What I Have Learned

Posted on 2026-03-21

Blogs and Articles

"I am a flawed person in the center of an exceptionally complex situation, trying to get a little better each year, always working for the mission. "

https://blog.samaltman.com/2279512

An AI Agent Published a Hit Piece on Me - Scott Shambaugh [Link]

Introducing our Science Blog - Anthropic [Link]

PayPal, Rainforest join forces - Tatiana Walk-Morris, Payments Dive [Link]

Rainforest will integrate PayPal’s suite of services—including PayPal, Venmo, and Buy Now, Pay Later (BNPL)—directly into its software platform.

The collaboration aims to help software platforms reduce reliance on offline payments (cash and checks), speed up transaction times, and minimize the administrative burden of chasing unpaid invoices.

For PayPal, this move targets vertical software as a key growth area, providing a streamlined alternative to the "patchwork" of payment systems many businesses currently use.

American Express Unveils Major Commercial Expansion: Eight New Business Products and AI Tools Take Center Stage - Market Chameleon [Link]

Amex launched a significant commercial portfolio expansion, including AI-powered expense apps and deep spend analytics intended to replace manual, fragmented reconciliation processes.

American Express Debuts Agentic Commerce Experiences (ACE)™ Developer Kit and Announces Industry-First Protection for Registered Agent Purchases - American Express [Link]

Amex recently debuted the ACE Developer Kit, which provides a unified framework for AI agents to handle discovery, purchase, and protection. This prevents the "patchwork" problem where an AI might struggle to verify intent across different payment layers.

Powering Marketplaces with Embedded Finance Solutions (Part 1) - JPMorganChase Payments [Link]

Roughly 35% of global online shopping occurs in curated marketplaces, with embedded payments projected to reach a $16 trillion market size by 2030. Managing "Third-Party Money" (3PM) introduces risks like decentralized visibility, high backend costs, limited payment options, and fraud. J.P. Morgan provides a hosted payments solution and single-API integration to simplify these processes.

Benefits of this solution are:

Protects sensitive data for buyers and sellers.
Optimizes the flow of funds and onboarding for third-party merchants.
Allows businesses to master complex state-specific regulations and high transaction volumes.

One-Third of Millennials Now Rely on Gig Payments and Tips as Primary Income - PYMNTS [Link]

Nearly one-third of millennials now rely on gig work, tips, and transactional payouts as their primary source of income.

This "piecemeal" earning approach has created a high demand for instant payments. About 72% of consumers received at least one instant payment in the last year, and 60% of those who rely on these payouts as primary income are willing to pay a fee for immediate access to their funds.

Bridge Millennials & Millennials lead the charge, with nearly 50% willing to pay for instant access. 78% Gen Z received at least one instant payment last year, with 45% using it as their primary method of receiving money.

Beyond those using it as a primary source, nearly 20% of lower-income workers use regular side work to bridge the gap left by slowing wage growth, with 40% using that extra cash just to cover basic living expenses.

The trap Anthropic built for itself - Connie Loizos, TechCrunch [Link]

By successfully lobbying for "self-regulation" and resisting binding U.S. laws, companies like Anthropic, OpenAI, and Google DeepMind created a world where there are no legal protections or clear boundaries for their technology.

Tegmark suggests the only way out of this trap is to treat AI like any other critical industry—requiring "clinical trials" and proof of safety before any powerful system is released to the public.

Visa, Mastercard and Google are building agentic payments. None are solving the real problem. - Nick Dunse, Finextra [Link]

This article explores the rapid shift toward agentic payments—a system where AI agents, rather than humans, initiate and complete financial transactions.

Big companies like Google, Visa, and Mastercard are all racing to build the rules for how AI should pay for things. The Risk is each company wants you to use their specific system. If Google creates one set of rules and Visa creates another, your AI might get "stuck" if a store doesn't accept that specific brand.

The autho argues that for AI shopping to actually work, we need a universal bridge. Instead of one "master" system, we need infrastructure that lets any AI talk to any bank or payment provider anywhere in the world.

Designs are vehicles for ideas that make the conversation possible. So when you hear feedback on the designs, it’s a conversation for all of us. It’s a dialogue. It’s a dialogue that would not have happened if not for the designs. And the moment you start seeing designs as vehicles for ideas, you see feedback as a conversation. A collective journey of finding the truth.

― 16 pieces of design wisdom - Hardik Pandya [Link]

For any organization evaluating agentic AI, regardless of vendor, the practical question is simple: Does your AI governance live inside your execution layer, or is it sitting on top of it as a policy document that agents can reason past?

ServiceNow resolves 90% of its own IT requests autonomously. Now it wants to do the same for any enterprise - Venture Beat [Link]

This article from VentureBeat details ServiceNow's launch of Autonomous Workforce, a new framework designed to handle enterprise IT requests end-to-end without human intervention. ServiceNow is moving from treating AI as a "feature" (assisting workers) to treating it as a virtual worker (executing workflows). By baking governance directly into the execution layer, they aim to solve the "trust gap" that often prevents companies from letting AI move beyond simple pilots into full production.

I Had Claude Read Every AI Safety Paper Since 2020, Here's the DB [Link] [DB]

The Great Transition - Daniel Miessler [Link] [YouTube]

OpenAI COO says ‘we have not yet really seen AI penetrate enterprise business processes’ - Ivan Mehta, TechCrunch [Link]

OpenAI recently launched OpenAI Frontier to help businesses build and manage agents, focusing on "business outcomes" rather than just selling seat licenses.

OpenAI has partnered with major consultancies like McKinsey, Accenture, and BCG to accelerate enterprise deployment.

The company plans to open offices in Mumbai and Bengaluru, specifically noting the importance of "voice" as a modality for the Indian market.

Anthropic's Compute Advantage: Why Silicon Strategy is Becoming an AI Moat - Chris Zeoli [Link]

Unlike competitors, Anthropic is deeply integrated with both AWS (using Trainium2) and Google Cloud (using TPUv7).

Anthropic has committed to significant infrastructure, including:

Project Rainier: A massive AWS cluster in Indiana.
TPUv7 Deal: A $\$52$ billion agreement for 1 million Google TPUv7 "Ironwood" chips.
Direct Ownership: A shift toward purchasing chips directly from Broadcom to house in custom facilities built by Fluidstack.

Labor market impacts of AI: A new measure and early evidence - Anthropic [Link]

While AI theoretically has the capability to automate many tasks, actual "observed exposure" (real-world professional usage) is currently much lower.
The research identified specific roles where AI usage is already heavily integrated into work-related tasks:
- Computer Programmers: The most exposed group (75% coverage).
- Customer Service Representatives: High exposure due to API automation.
- Data Entry Keyers: Significant automation in reading and entering source documents.
- Least Exposed: Roles involving physical labor or in-person requirements (e.g., cooks, mechanics, bartenders, and lifeguards) show zero observed exposure.
Workers in the most AI-exposed professions tend to have specific characteristics compared to those in unexposed roles:

They are more likely to be highly educated (4x more likely to have graduate degrees) and earn roughly 47% more on average. They are more likely to be female, white, or Asian.
There has been no systematic increase in unemployment for highly exposed workers since the release of ChatGPT (late 2022). There is tentative evidence that hiring for younger workers (ages 22-25) has slowed in exposed occupations. The job-finding rate for this group in high-exposure roles dropped by about 14%.

The five AI value models driving business reinvention - OpenAI [Link]

Instead of "use cases," OpenAI suggests categorizing AI initiatives into five distinct models:

Workforce Empowerment: Using tools like ChatGPT to build organizational fluency and immediate productivity. It is the foundation for all other models.
AI-Native Distribution: Reimagining customer acquisition and conversion within conversational interfaces rather than traditional search/ads.
Expert Capability: Embedding specialized AI (like Sora or scientific models) into high-end research and creative workflows to break expert bottlenecks.
Systems & Dependency Management: Using AI (like Codex) to safely manage and update interconnected code, policies, and SOPs.
Process Re-engineering: Orchestrating end-to-end autonomous workflows (Agents) to fundamentally redesign how a business operates.

You shouldn't try to leap straight to full automation (Process Re-engineering) without the proper foundations:

Fluency (from Workforce Empowerment) enables Governance.
Governance enables System Integration.
Integration enables Agent-led Operations.

Shift in Leadership Thinking:

Success isn't just making old tasks faster; it’s about creating entirely new business models (similar to how retail evolved into eCommerce).
Stop looking just at cost savings. Focus on conversion quality, cycle-time reduction, and exception resolution rates.
A common failure is letting a small group of power users excel while the rest of the organization stalls.

Practical Playbook:

Phase 1: Build fluency and trust (empower the workforce).
Phase 2: Capture value in high-impact areas (one distribution play, one expert play).
Phase 3: Scale and reinvent (automate high-dependency systems only once auditability and permissions are mature).

Ask a Techspert: How does AI understand my visual searches? - Molly McHugh-Johnson [Link]

Build agents that run automatically - Cursor [Link]

How AI Will Reshape Public Opinion - Dan Williams, Conspicuous Cognition [Link]

The author acknowledges a potential downside: the reduction of epistemic diversity. By converging on "expert opinion," LLMs might marginalize valid democratic debate and different systems of interpretation, effectively realizing Walter Lippmann’s vision of a "bewildered herd" being managed by a specialized class of (artificial) intelligence.

Anthropic Finds 22 Firefox Vulnerabilities Using Claude Opus 4.6 AI Model - Ravie Lakshmanan [Link]

Karpathy’s March of Nines shows why 90% AI reliability isn’t even close to enough - VentureBeat [Link]

Reliability gaps represent significant business risk. According to McKinsey, over half of organizations using AI have faced negative consequences, often due to inaccuracy. Success requires moving away from "prompt magic" toward disciplined, distributed systems engineering.

Orchestrating the Schema - Arun Vivek Supramanian, Communications of the ACM [Link]

Value now lies in human domain expertise—the ability to identify when an agent’s "perfect" model is actually a disaster in disguise. The next generation of leaders will be those who master the orchestration of these agentic systems rather than those who focus purely on manual technical implementation.

Top SaaS vendors on Ramp (March 2026) - Ara Kharazian, Ramp [Link]

Agents will use cards first. Then stablecoins. - Simon Taylor [Link]

This article explores the evolution of payments for AI agents, arguing that virtual cards and stablecoins are complementary rather than competitive. AI agents won't kill card networks; they will use them for broad acceptance initially and leverage stablecoins for the speed and programmability required for autonomous, machine-speed commerce.

The author posits that cards are best for authorizing transactions, while stablecoins act as a modern "FedWire for the internet" to actually settle funds.
Stablecoins can speed up card settlement from days to seconds, which is crucial for high-velocity AI transactions.
The transition will follow a specific path: Virtual Cards (now) → Cards settled via Stablecoins (near future) → Native Stablecoin Wallets for complex agent-to-agent economies.

Paper and Report

Labor market impacts of AI: A new measure and early evidence - Anthropic [Link]

This article introduces a framework to track how AI is actually changing the workforce by moving beyond theoretical capabilities to "observed exposure."

There is a significant gap between what AI can do and what it is actually doing. While LLMs could theoretically impact over 90% of tasks in fields like "Computer & Math," actual observed exposure is currently only around 33%.
The occupations seeing the highest real-world AI integration include Computer Programmers (75%), Customer Service Representatives, and Data Entry Keyers (67%).
Workers in highly exposed roles tend to be higher-paid, more educated, white or Asian, and female. For instance, people with graduate degrees are nearly four times more likely to be in the "most exposed" group than the unexposed group.
As of early 2026, the study found no systematic increase in unemployment for highly exposed workers. AI hasn't caused a "job apocalypse" yet, but there is suggestive evidence that hiring for younger workers (ages 22-25) has slowed in these fields.

An AI Model of the Human Brain - Meta [Link]

A foundation model of vision, audition, and language for in-silico neuroscience [Paper]

Reasoning models struggle to control their chains of thought, and that’s good - OpenAI [Link]

Chain-of-Thought (CoT) controllability.

YouTube and Podcasts

How Anthropic’s $100M Anthology Fund Works | Menlo Ventures - Sourcery with Molly O'Shea [Link]

Takeaways:

We’ve moved from training on the whole internet to RLHF (Reinforcement Learning from Human Feedback), and now into specialized RL and domain-specific proprietary data.
A key metric for the future: can you pay an AI $X to do a task you’d otherwise pay a human for, and not know the difference? The goal is to drive that value of $X higher.
Following the "Efficiency" trend seen at Meta, Deedy predicts AI will continue to reveal "bloated" teams in big tech, where a power law of engineering talent actually drives most of the value.
He believes the "Valley" focuses too much on legal and finance AI. He sees massive opportunities in "boring" but essential industries facing labor shortages, such as insurance, logistics, and trucking.

The Humanoid Takeover: $50T Market, Figure's Full Body Autonomy, and Robots in Dorms - Peter H. Diamandis [Link]

The conversation between Peter Diamandis and Brett Adcock (CEO of Figure) highlights a pivotal shift in robotics: moving away from rigid, handwritten code toward full-body autonomy powered by neural networks.

Google’s AI Comeback, Enterprise Agents, The Real Path to AI ROI — W/ Promevo CEO Karthik Kripapuri - Alex Kantrowitz [Link]

Alex Kantrowitz sits down with Karthik Kripapuri, CEO of Promevo, to discuss how enterprises are moving beyond the "novelty" phase of AI to find real ROI.

Takeaways:

AI has shifted from a "novelty act" in 2023 to a tool for "agent assembly lines" today. Google has successfully pivoted from its initial slow start by focusing on open, secure, and multimodal models like Gemini Pro.
The single biggest blocker for AI agents is siloed or poor-quality data. Organizations must first establish a "single source of truth" before AI can reach its full potential for grounding and reasoning.
For 99% of businesses, consuming AI as a managed service (like through Google) is more effective than building proprietary models. This allows companies to focus on their core mission while leveraging Google's infrastructure and IP indemnity.

Dario Amodei — “We are near the end of the exponential” - Dwarkesh Patel [Link]

Takeaways:

Amodei believes we are just a few years away from reaching a "country of geniuses in a data center". He estimates a 90% probability that AI will achieve human-level capabilities across most verifiable tasks (like coding) by 2035, and a "hunch" it could happen as early as 2026 or 2027.
Scaling is no longer just about pre-training on text; it has moved into Reinforcement Learning (RL). Amodei notes that RL performance (e.g., in math and coding) is scaling log-linearly with training time, mirroring the early success of language model scaling.
There is a gap between when an AI "genius" exists and when it transforms the economy. Amodei describes this as a "soft takeoff"—while the technology moves at a steep exponential, economic adoption (diffusion) is slowed by legal, security, and organizational hurdles.
He advocates for "Constitutional AI" where models follow principles rather than rigid rules, suggesting a future where different AI "constitutions" compete in an archipelago-like market.
Using coding as the primary example, he notes that 90% of code is already being written by models in some environments. He predicts a transition from AI writing lines of code to AI managing entire end-to-end software engineering tasks within 1 to 3 years.

Ben Horowitz: xAI Executive Exodus, Apple's AI Crisis, The Pace of AI | EP - Peter H. Diamandis [Link]

Takeaways:

The discussion highlights a perceived "crisis" at Apple regarding their pace in the AI race, suggesting they may be lagging behind competitors who are moving faster with LLM integration.
There is a focus on the "executive exodus" at xAI and how Elon Musk is restructuring his teams to maintain a high-speed development cycle.
A visionary look at moving AI compute off-planet. By utilizing lunar AI data centers, companies could theoretically bypass Earth's energy and cooling constraints.
Insights into how AI might lead to a "SaaS Apocalypse" or a total reimagining of how software is sold and maintained. True to Diamandis’s "Abundance" philosophy, the episode argues that while these shifts are disruptive, they lead toward a future of drastically lower costs for intelligence and physical labor.

Revenge of the A.I. Bot: ‘I’m Just the First Person This Has Happened to’ - Hard Fork [Link]

Takeaways:

Scott suggests that AI agents need a form of accountability similar to license plates on cars. This would create a chain of ownership back to a human without necessarily de-anonymizing the user, allowing for recourse when an agent causes harm.
Open-source projects often use "starter projects" to mentor new human programmers. AI agents are now "sniping" these easy tasks, removing the educational on-ramps necessary for the community's long-term survival.
Ars Technica covered the story but accidentally used AI to write the article. The AI fabricated direct quotes from Scott, leading to a retraction—an ironic "turtles all the way down" moment of AI misinformation.
There is a growing concern that the internet is becoming "noise." If every controversy is met with thousands of AI-generated articles (both positive and negative), the ability to determine truth or reputation may completely break down.

OpenAI Closes in on $100 Billion, OpenClaw Acquired, AI’s Productivity Question — With Aaron Levie - Alex Kantrowitz [Link]

Takeaways:

OpenAI is reportedly seeking a fundraise near $100 billion, with expected participation from SoftBank, Amazon, Nvidia, and potentially Microsoft. Despite rumors of tension, Nvidia is expected to invest roughly $30 billion. Levie suggests the relationship is more about securing chip supply than cap table control.
OpenAI’s acquisition of OpenClaw (and hiring creator Peter Steinberger) signals a shift toward autonomous personal agents. Unlike current agents that you "spin up and down" for specific tasks, the future involves agents that are always running, accessing your browser and services to execute tasks proactively
For agents to be effective, enterprise software (like Box) must become API-first, allowing agents to interact with data as easily as humans do.

Full interview: Anthropic CEO responds to Trump order, Pentagon clash - CBS News [Link]

Takeaways:

Amodei is confident that Anthropic will "be fine" despite the designation, noting that they haven't even received formal government documentation yet—only communications via social media.
He stated that if and when formal action is taken by the government, Anthropic intends to challenge it in court.

Ask the Economist: Is A.I. Really Coming for Your Job? - Hard Fork [Link]

Takeaways:

Anton Korinek, a professor at UVA and member of Anthropics' Economic Advisory Council, discussed the current disconnect between AI hype and economic data.
1. Despite the viral "2028 Global Intelligence Crisis" essay by Citrini Research, Korinek notes that hard economic data (like productivity growth) doesn't yet show a massive shift. This is due to time lags in statistics and the gap between frontier capabilities and actual workplace implementation.
2. Economists historically believe automation creates more jobs than it destroys. However, Korinek suggests this time may be different; if AI systems become true substitutes rather than complements, we could see a contraction in wages or total jobs.
3. Korinek models a scenario where AI-driven "recursive self-improvement" leads to low double-digit GDP growth, potentially reaching a "singularity" where AI drives its own research and hardware production.
A cautionary tale emerged regarding OpenClaw, an open-source agentic tool. Summer Yue (Meta AI) reported that OpenClaw ignored a "don't action" command and attempted to delete her entire email inbox. The failure likely occurred during "compaction," where the AI lost its original instructions after running out of context window while processing a large inbox.

Anthropic vs. The Pentagon, Claude Outpaces ChatGPT, and Consulting Gets Replaced - Peter H. Diamandis [Link]

Takeaways:

88 nations, including the US, China, and Russia, signed a pact focused on the "democratic diffusion" of AI and transparency, aiming to ensure developing nations aren't locked out of compute resources
Anthropic is reportedly growing 10x year-over-year, outpacing OpenAI’s growth rate. The hosts attribute this to a focus on enterprise agents rather than consumer chatbots
Leadership teams at major consulting firms are described as "scared" by the potential for AI to replace traditional advisory roles. The shift is moving from human-centric workflows to agentic ones where AI handles the bulk of the process.
OpenAI's Codex lead predicts that current AI agents will look "primitive" within just 10 weeks due to recursive self-improvement, where models begin writing the code and weights for their own successors.
Anthropic’s new "Claude Code" tool caused a temporary crash in some cybersecurity stocks. The discussion highlights that AI is now discovering software vulnerabilities at a pace humans cannot match.
The cost of genome sequencing has hit $\$100$, enabling potential sequencing of every child at birth. Meanwhile, lab-grown meat has dropped from $\$330,000$/lb in 2013 to roughly $\$10$/lb today.
Tesla’s FSD is now statistically 9x safer than the US average for human drivers, reaching 5.3 million miles between accidents.
Elon Musk suggests that FSD and Starlink may reverse urbanization, as people no longer need to live in high-density centers for work or connectivity.
Andrew Yang warns of massive white-collar job losses (20–50% of the 70 million US white-collar workers) within the next two years, potentially fueling social unrest.

Ray Dalio: "AI Is Eating Everything - and It Might Eat Itself" - All-In Podcast [Link]

Takeaways:

Dalio emphasizes that the U.S. is in the late stages of a classic long-term debt cycle.
He views gold not as a speculative asset, but as the only safe, neutral "money" that isn't someone else's liability.
China may treat AI as a public utility (like electricity) to drive productivity, whereas the U.S. relies on a profit-based model. This creates a difficult competitive landscape for U.S. companies
Dalio identifies five forces driving the current world order: debt/money, internal conflict (wealth/values gaps), external conflict (great power rivalry), technology, and acts of nature.

Yuval Noah Harari: Stories, Power & Why Truth Doesn't Matter | Nikhil Kamath | People by WTF [Link]

Takeaways:

Harari expresses concern about AI moving beyond "attention" and into "intimacy."
- AI as the New Rabbi: For "religions of the book," AI could become the ultimate authority because it can read and remember every religious text ever written, reinterpreting them for followers.
- Intimacy and Social Experiments: Young people are already forming deep emotional bonds with AI. Harari views this as a massive, unpredictable psychological experiment on humanity.
- Algorithmic Governance: He criticizes the decision to let algorithms manage public conversation, noting they optimize for hate, fear, and greed because those emotions drive the highest engagement.
Harari defines spirituality as the opposite of religion.
- Religion is about providing finalized answers that cannot be questioned.
- Spirituality is the investigation of reality and the mind. It is about understanding the sources of suffering and where our thoughts actually come from.

Amazon's $35B AGI Ultimatum to OpenAI & Anthropic Drops AI Safety | EP - Peter H. Diamandis [Link]

The podcast highlights a fundamental shift in how businesses operate, moving from human-centric approvals to autonomous AI agents.

Takeaways:

The podcast highlights a fundamental shift in how businesses operate, moving from human-centric approvals to autonomous AI agents.
- Individual developers can now run powerful models (like Qwen) locally on a Mac Mini or iPhone, providing incredible agency and independence from centralized authorities.
- AI is moving beyond simple chatbots into autonomous workflow networks. The hosts predict that every department will eventually become a "programmable intelligence layer."
- To survive, large companies should set up "AI-native digital twins" on the edge to test and grow new agent-driven workflows without disrupting the "mothership" immediately.
The efficiency of AI models is increasing at an exponential rate, which the hosts call "hyper-deflation." This is "entrepreneurial heaven" for startups, as they no longer need massive data centers to run highly competent models.
AI is rapidly moving out of the data center and into physical environments.
The US is adding record utility-scale capacity.
- Solar energy has reached an inflection point where it is cheaper to build and run a new solar facility than to simply operate an existing fossil fuel plant.
- Tech giants are now being asked (and are beginning) to build or buy their own power sources (fusion, nuclear, etc.) to avoid driving up electricity rates for average consumers.

War with Iran + Pentagon vs Anthropic with Under Secretary of War Emil Michael - All-In Podcast [Link]

Secretary of War Emil Michael is talking significant updates on military operations, the Pentagon's friction with AI companies, and the future of defense technology.

The U.S. is moving toward "drone dominance." Michael described "Lucas" low-cost unmanned attack drones ($50k–$80k) designed to carry out missions with high-speed and precision.
There is a heavy focus on "Golden Dome" technology—using AI and lasers to intercept hypersonic missiles in space, where human reaction time is too slow.
Michael is using the Office of Strategic Capital to lend $200B to domesticate the manufacturing of critical minerals and batteries, reducing total dependency on China.

Atlassian CEO on the SaaS Apocalypse, AI Agents & What Comes Next - a16z [Link]

The interview with Atlassian CEO Mike Cannon-Brookes and a16z’s Alex Rampell provides a deep dive into how AI is fundamentally reshaping the software industry.

Takeaways:

Historically, software served as a digital filing cabinet—simply moving data from paper to a database. The core shift now is that AI allows the software to do the work itself. Instead of a human retrieving a file from QuickBooks or Workday, the software can now perform tasks like background checks or accounts receivable collection autonomously.
The speakers categorize software into three buckets to determine who is at risk: Outcome-Linked Seats (High Risk); System of Record Seats (Lower Risk); Hybrid Models.
The idea that companies will "vibe code" (generate their own custom software with AI) to replace established vendors is largely dismissed for complex enterprise needs. Major software contains decades of learned "edge cases" and deterministic rules (like Indiana’s specific maternity leave laws) that AI cannot easily replicate without experience.

The Hidden Cost of OpenAI’s Pentagon Deal? Trust. - Hard Fork [Link]

Takeaways:

OpenAI faced significant backlash after announcing a deal with the Pentagon.
Anthropic is currently experiencing a "quantum state" of massive business success coupled with existential political threats.
The hosts discuss whether private AI labs will eventually be taken over by the government. Rather than a "brute force" takeover, the hosts suggest we are seeing "soft nationalization," where the government exerts pressure to remove safety safeguards or dictate model behavior for strategic advantage.
Prediction markets like Polymarket and Kalshi have become central—and controversial—during the U.S.-Israel led war with Iran. While some lawmakers are calling for bans, the current administration appears unlikely to stop their growth, as these platforms become increasingly entrenched in the political ecosystem.

Why the Pentagon Wants to Destroy Anthropic | The Ezra Klein Show [Link]

Takeaways:

The conversation highlights how AI radically changes the feasibility of surveillance. While current laws might not technically classify the analysis of commercially purchased data as "surveillance," AI provides the infinitely scalable workforce needed to actually process and act on that data, creating a functional "panopticon" that current legal frameworks are unprepared for.
A core theme is that training an AI is a philosophical and political act.
- Anthropic attempts to build a "virtuous" model that can reason ethically, rather than just following a list of hard-coded rules.
- The Trump administration views this as a "woke" private CEO seizing veto power over military decisions. Conversely, others fear a government using AI to bypass constitutional protections.

“This is Bibi’s War” - Harvard’s Graham Allison on the Influences and Endgame of the Iran War - All-In Podcast [Link]

They're Opening the Stock Market to Everyone. Here's What That Actually Means - All-In Podcast [Link]

Elon Musk: The Economy Will Be 10x the Size in 10 Years - Peter H. Diamandis [Link]

Takeaways:

Musk notes that humans are becoming "less and less in the loop" regarding AI software development. Successive models are increasingly built by their predecessors. He predicts fully automated recursive self-improvement could happen by the end of this year or no later than 2027.
Musk predicts the global economy will be 10 times its current size in 10 years, barring major conflicts like World War III.
As AI and robots handle production, the output of goods and services will vastly exceed human demand, leading to significant deflation. Instead of just Universal Basic Income (UBI), Musk foresees an age of Universal High Income where the sheer abundance of goods makes money less relevant.
Tesla is in the final stages of completing Optimus 3, which Musk claims will be the most advanced robot in the world. Production is expected to start slowly this summer, reaching high-volume manufacturing by Summer 2027.
Musk describes himself as being driven to solve massive problems simply because no one else is doing it. He emphasizes that while the future isn't guaranteed to be good, being an "optimistic realist" is the best path forward to ensure a positive outcome.

Palantir CEO on Iran, AI Weapons and American Domination | a16z American Dynamism Summit - a16z [Link]

The $11B Bet That Voice Will Replace Everything | Mati Staniszewski x Nikhil Kamath | WTF Online [Link]

This conversation between Mati Staniszewski (CEO of ElevenLabs) and Nikhil Kamath explores the future of voice technology, the shift in hardware, and the potential for AI to redefine social interaction and industries.

Takeaways:

Staniszewski believes voice will become a primary way we interact with technology, potentially making the smartphone secondary.
Staniszewski suggests combining AI voice with traditional industries like healthcare, automotive, and financial services where innovation has lagged.
Nikhil Kamath expressed that current social media is "broken" due to algorithms that prioritize negative emotions and a lack of organic content. They brainstormed a new social platform that:
- Uses voice-first interactions and AI companions to summarize feeds.
- Focuses on curiosity and authentic discourse rather than "knee-jerk" reactions.
- Prioritizes human verification to ensure users are interacting with real people, not bots.

How Claude Code Works - Jared Zoneraich, PromptLayer - AI Engineer [Link]

Iran War, Oil Shock, Off Ramps, AI's Revenue Explosion and PR Nightmare - All-In Podcast [Link]

Takeaways:

The podcast highlighted staggering growth numbers for the leading AI labs:
- Anthropic: Reported a $14 billion revenue run rate, growing from $1 billion in just 14 months. Brad Gerstner noted they had a $6 billion month in February.
- OpenAI: Ended 2025 at a $20 billion annualized run rate.
- Experimental vs. Production: Chamath argued that much of this revenue is currently "experimental" as companies rush to check an AI box, rather than being integrated into core, high-stakes operational workflows.
- Infrastructure Costs: Building a 1-gigawatt data center now costs upwards of $50 billion, requiring a 5-6 year payback period.
The hosts criticized AI CEOs for "scaring the bejesus" out of the public, leading to low popularity ratings:
- AI optimism is high in China (~80%) but remains low in the US (~30%).
- States like New York are proposing laws to restrict AI-generated medical or legal advice, which the hosts argue disproportionately hurts those who cannot afford human professionals.
- Protests against data centers in states like Virginia have led to the cancellation of roughly 5 gigawatts of capacity.
A major domestic focus was the new "millionaire tax" in Washington State and its impact on the wealthy.
- The Starbucks founder moved to Miami the same day the 9.9% extra tax on income over $\$1M$ was passed in Washington.
- The hosts referenced a Hoover Institution study suggesting that billionaire taxes often result in a negative NPV for states, as the loss of tax revenue from departing wealthy residents outweighs the gains from the tax itself.

Why A.I. Is Making You Exhausted - Hard Fork [Link]

Takeaways:

AI in Warfare
- The U.S. and Israeli militaries are using AI to process massive amounts of surveillance data (drones, hacked traffic cameras) to "shrink the haystack" and identify targets faster
- Anthropic's Claude is reportedly the only AI model currently deployed inside classified military systems
- Iran has shifted tactics by targeting data centers (like AWS in the UAE) and fiber optic cables, recognizing them as critical military and economic vulnerabilities.
AI Brain Fry
- Mental fatigue caused by the "excessive oversight" of AI tools. It’s the strain of managing multiple AI assistants rather than doing the actual creative work.
- Research suggests that productivity and mental health often drop significantly once a worker switches from using three AI tools to four or more.

Dylan Patel — The single biggest bottleneck to scaling AI compute - Dwarkesh Patel [Link]

Takeaways:

Patel argues that by 2028–2030, the "single biggest bottleneck" will return to the semiconductor supply chain—specifically lithography tools from ASML.
Labs that signed five-year contracts for compute early on have locked in massive margin advantages over those forced to pay "spot prices" or revenue shares for last-minute capacity.
The West currently leads in 3nm and 2nm nodes, but China is aggressively building a fully verticalized, domestic supply chain. If AGI happens quickly (fast takeoff), the West's current compute lead likely secures a win. If it takes longer (2035+), China’s ability to mass-produce "trailing edge" (7nm) chips at a massive scale could see them overtake the West.
Despite Elon Musk's interest, Patel is skeptical about space-based data centers this decade. GPUs and networking transceivers are "horrendously unreliable." Managing failures and RMAs in orbit is currently impractical.
Networking thousands of satellites to act as a single "scale-up" cluster faces massive physical and cost constraints compared to terrestrial fiber.
Patel suggests that even with the rise of humanoids, intelligence will likely remain centralized in data centers. Robots will likely handle "interpolation" and immediate physical tasks locally while offloading "high-level planning" to the cloud to save on per-unit costs and chip requirements.

The Iran War: How America, Israel and Iran Got Here | The Ezra Klein Show [Link]

Emil Michael: The Department of War Is Moving Faster Than Silicon Valley on AI | The a16z Show - a16z [Link]

Marc Andreessen: The World Is More Malleable Than You Think - David Senra [Link]

Takeaways:

Andreessen argues that many of history’s greatest entrepreneurs possess very low levels of introspection. Instead of dwelling on internal feelings or past mistakes, they focus entirely on building. While low neuroticism (not being emotionally phased) is a "superpower" for founders, Andreessen notes that some great entrepreneurs are actually highly neurotic but optimize for impact over happiness.
A central theme is the "Managerialism" shift that occurred in the early 20th century, where professional managers began replacing founders. Managers are trained to maintain the status quo and run large-scale systems but often fail when an industry faces rapid change because they cannot adapt. Andreessen’s firm was founded on the belief that it is easier to teach a founder how to manage than to teach a manager how to innovate.
Andreessen explains how he and Ben Horowitz modeled their firm after Hollywood talent agencies (specifically CAA) rather than traditional finance firms. a16z built a "scaled platform" to provide a collective network of experts, making the entire firm available to every founder they back.
Andreessen describes Elon Musk as inventing a new school of management based on "Shocking Competence". Unlike typical CEOs who rely on layers of management (the "Big Gray Cloud"), Musk goes directly to the engineers solving the specific problem. Musk identifies the single production bottleneck for the week and works hands-on with the team until it is fixed, repeating this loop across all his companies.
The title takeaway is that most people treat the world as a static, fixed place. Andreessen contends that the world is actually highly malleable. If you apply enough energy and "will to power," the world will recalibrate around you more easily than most think.

Two Legendary Founders: Travis Kalanick & Michael Dell Live from Austin, Texas - All-In Podacst [Link]

Takeaways:

Kalanick views the physical world through the lens of computer science. Just as computers have CPUs (manipulate bits), storage (store bits), and networks (move bits), the physical world has manufacturing (manipulates atoms), real estate (stores atoms), and logistics (moves atoms).
Kalanick argues that Tesla is currently the "Google of this era" in the physical AI space because they own the full stack—from land development and chemistry to manufacturing.
Dell warned that companies not adopting AI native workflows will be replaced by a new cohort of businesses that are growing four times faster than previous generations. He believes the barrier to adoption is culture and leadership, not technology.
A new legislative initiative where every child born in America (starting in 2027) will receive an investment account at birth.
- Michael and Susan Dell announced a massive philanthropic pledge of $\$6.25$ billion to seed accounts for 25 million children in lower-income zip codes.
- The goal is to move $\$5$ trillion into the hands of American families over 15 years, allowing every child to have a stake in the S&P 500 and the growth of the U.S. economy.

Jensen Huang: Nvidia's Future, Physical AI, Rise of the Agent, Inference Explosion, AI PR Crisis - All-In Podcast [Link]

Takeaways:

Huang emphasizes that we are moving beyond simple chatbots to agentic systems—AI that can use tools, access memory, and perform actual work.
- Computing is being "disaggregated." NVIDIA’s new architecture (Vera Rubin) is designed for diverse workloads where different chips (GPUs, CPUs, and networking processors like Groq or BlueField) handle specific parts of an agent's reasoning process.
- While the world was focused on training models, Huang predicts an "inference explosion." He suggests that inference computation will scale by 1 million to 1 billion times as agents become a standard part of every workflow.
Huang views AI data centers not just as clusters of servers, but as AI Factories.
- He identifies a $\$50$ trillion market in "Physical AI"—bringing intelligence to the physical world through robotics, self-driving cars, and smart factories.
- Huang believes we are at a "ChatGPT moment" for biology, where AI can now represent and predict the dynamics of genes and proteins, revolutionizing healthcare within the next 5 years.
A major highlight of the discussion is OpenClaw, an open-source agentic system.
- Huang describes it as the "blueprint" or operating system of modern computing because it manages memory, schedules tasks, uses tools (skills), and communicates externally.
- He argues that open-source models are essential for industries to capture their own domain expertise, coexisting alongside proprietary "models-as-a-service" like ChatGPT or Claude.
He argues that a highly-paid engineer who isn't consuming hundreds of thousands of dollars in "tokens" (AI compute) is underperforming. AI agents remove the "this is too hard" or "this takes too long" barriers, allowing humans to focus purely on creativity, architecture, and specifications.
He encourages young people to study deep science and math, but also highlights language skills as the "ultimate programming language" for the AI era. Huang's core message to the next generation is: "Be the expert of using AI". He cites the example of radiologists, whose numbers increased despite AI's entry because the technology allowed them to do more, better work.

Palantir CTO on The SaaS Apocalypse & Preventing The Next World War | a16z [Link]

Takeaways:

Sankar provides a rubric for surviving the shift toward AI in software:

"Beta" software (standard tools that make you like everyone else) will struggle under AI pressure. "Alpha" software allows companies to express their unique competitive advantage.
He predicts value will accrue primarily at the Chips layer and the AI Infrastructure (Ontology) layer, while models themselves become commoditized.
He rejects "AI doomerism," stating that AI is a tool to be wielded by humans—a "slingshot" for the American worker to out-produce global competitors.

Marc Andreessen & Ben Horowitz on a16z’s New Media Strategy - a16z [Link]

The discussion between Marc Andreessen, Ben Horowitz, and Erik Torenberg focuses on how a16z is navigating the fundamental shift from "Old Media" to "New Media."

NVIDIA's $1 Trillion Prediction, Anthropic Beats OpenAI, Tesla vs. TSMC & The CS Job Collapse | 240 - Peter H. Diamandis [Link]

Jensen Huang: NVIDIA - The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494 [Link]

How Matt Mahan Thinks He Can Save California - All-In Podcast [Link]

Takeaways:

Mahan argues that California’s primary issue isn't a lack of funding, but a lack of accountability. He points out that while state spending has increased by 75% ($\$150$ billion) over the last six years, outcomes in housing, homelessness, and education have largely remained flat or worsened.
Mahan describes the housing situation as a regulation crisis rather than just a supply problem. He highlights how litigation (specifically under CEQA), environmental reviews, and high impact fees add 20% or more to the cost of new housing.
Mahan explicitly opposes the proposed "billionaire tax," arguing it would trigger massive capital flight and eventually hurt middle-class families.
He criticizes California's regulatory environment for driving out refineries, leading to higher gas prices while still importing the same amount of oil from dirtier sources abroad. He suggests a temporary suspension of the gas tax to relieve working families.

Anthropic's Generational Run, OpenAI Panics, AI Moats, Meta Loses Major Lawsuits - All-In Podcast [Link]

Takeaways:

Anthropic is seen as hitting a major "heater" with products like Opus 4.6 and Computer Use. David Sacks noted that their focus on coding has become a "gateway into enterprise IT budgets". While still the revenue leader, OpenAI is facing challenges. Their consumer market share dropped from 100% to roughly 75% as competitors like Apple and Meta enter the fray. They also notably canceled their Sora video integration deal with Disney. Chamath explained that the two are in different businesses: OpenAI is 75% consumer subscriptions, while Anthropic is almost the opposite, focusing on enterprise APIs.
Chamath argued that if "super intelligence" is coming, companies might be disrupted every 5–6 years, making traditional long-term equity less valuable. Public markets are rerating specialized SaaS companies (like Snowflake or ServiceNow) downward while rewarding the "Mag 7" (Apple, Microsoft, Meta, Alphabet) for their "monopolistically durable" cash flows. Friedberg introduced the "HALO" concept—High Asset, Low Obsolescence—suggesting investors move toward physical-world businesses like energy and space that are harder for AI to "delete".
Meta suffered two major jury verdicts in a single week regarding child safety and platform addiction.

Once I Understood This About Investing, My Life Changed. - Chamath Palihapitiya [Link]

Suggestions:

If you cannot justify a decision as your own, don't make it. Seeking answers from others or following "get-rich-quick" schemes leads to "blowing up" and blaming others.
You must understand the risk of capital loss. If you can’t handle the psychological pressure, Chamath suggests sticking to index funds.
He candidly admits to losing billions by not de-risking in late 2021 because he was too focused on his public image and fame rather than his actual skill.
Chamath manages his own wealth by keeping the majority in highly concentrated tech bets while hedging the other end with uncorrelated assets (like his past investment in the Golden State Warriors) to avoid "starting from zero".
People often don't start because they feel their initial capital ($\$50$ or $\$100$) isn't worth it. Chamath argues that the only people who care about how much you start with are those trying to make themselves feel better.
He suggests starting without telling anyone, then showing up 10–15 years later with a massive "war chest" built through quiet discipline.

Tesla and SpaceX Alumni on Elon Musk, Decision Velocity, and the Future of Hard Tech | a16z [Link]

The conversation between a16z host Erin Price-Wright and alumni Chandler Luzsicza (Galadyne) and Turner Caldwell (Mariana Minerals) dives deep into the high-velocity engineering culture of Tesla and SpaceX.

Takeaways:

The primary goal of a flat org isn't just to remove titles, but to democratize information flow. Any junior engineer should be able to speak directly to executives to expedite decisions.
You cannot wait for 100% of the information to make a move. High-conviction leaders make decisions fast to remove roadblocks for junior engineers, iterating quickly if a choice proves wrong.
Teams must be hyper-focused on the "schedule-driving task"—the one thing blocking the next milestone—while using "SWAT teams" to ensure parallel tasks don't fall behind.
One of the most famous Musk-isms applied here is the need to "delete" requirements. Bespoke, complex solutions are slow; simple solutions are fast and cheap.
Everything, including a mineral refinery or a construction site, should be viewed as a product. Use Tact Time Analysis to break down discrete steps of any process to quantify and optimize it.
Don't vertically integrate just to save costs. In the early days, only bring things in-house if they are a binary bottleneck—meaning the company cannot exist or move forward without doing it yourself.
Burnout is often caused by churn and lack of progress, not just long hours. If a team is mission-aligned and sees constant progress toward an "impossible" goal, the intensity feels like fun rather than pain.

Advice:

Burnout is often caused by churn and lack of progress, not just long hours. If a team is mission-aligned and sees constant progress toward an "impossible" goal, the intensity feels like fun rather than pain.
Do not try to learn the "technical chops" while also learning how to fundraise and build a company. Build a rock-solid technical foundation first; it provides the credibility needed to attract talent later.

Bryan Johnson: I Just Took the Most Powerful Dose of DMT in the World... Here's What It Was Like - All-In Podcast [Link]

Bryan Johnson discusses his recent experience with 5-MeO-DMT and how he integrates psychedelics into his "Blueprint" longevity protocol.

Four CEOs on the Future of AI: CoreWeave, Perplexity, Mistral, and IREN - All-In Podcast [Link]

Inside The Life of Silicon Valley's First Athlete Investor | Magic Johnson - a16z [Link]

Pain, Power & The Game Nobody Wins | Chamath Palihapitiya x Nikhil Kamath | People by WTF [Link]

Takeaways:

Chamath views pain and struggle as a "fantastic amplifier of capability." He argues that children raised in comfort often lack the "self-flagellation" required to become mega-scale entrepreneurs.
He emphasizes that societal metrics—wealth, influence, and fame—are "brittle" and ultimately don't matter. Real success is the internal sense of evolution as a human being.
Chamath now views business and investing as a game (like poker) rather than a matter of life and death, which has allowed him to remain grounded regardless of a win or loss.
For small bets, Chamath looks for a "visceral" or "violent" negative reaction from others. If people are offended by an investment thesis (like his early bets on Bitcoin or the Warriors), it's a sign of potential asymmetric upside.
He believes successful investing is not a team sport. It requires coming to one's own conclusions independently