NEW RULES FOR FOUNDERS

The startup game has changed.

AI increases individual leverage, code becomes cheaper, and small teams can operate with capabilities that once required departments.

The founder advantage now comes from judgment, agency, direct knowledge, reusable systems, and the courage to follow truths others ignore.

FOUNDER MINDSET

THE FIRST MISTAKE: CHASING WHAT WAS HOT

After the dot-com crash, web work looked dead. Garry moved toward Windows Mobile because mobile appeared to be the next hot thing.

In hindsight, he already had years of web experience just before social software exploded.

Lesson: abandoning your accumulated edge to chase consensus can be expensive.

EARNESTNESS REQUIRES COURAGE

Being earnest is not simply being naive.

It means being willing to say: “My direct experience tells me something different.”

A blog post, investor, trend report, or popular opinion should not automatically override what you know from customers, products, technical work, or repeated first-hand evidence.

STOP ASKING: WHAT IS HOT?

The wrong question:
“What should I work on because it is hot?”

Better questions:
1. What am I genuinely interested in?
2. What do I know unusually well?
3. What problem have I seen directly?
4. Where do I have years of accumulated context?

Your edge often starts there.

MAP VS. TERRITORY

Garry describes another major mistake: saying no to joining the people who became Palantir because he was focused on career status and promotion.

He was reading the map: titles, prestige, what looked smart.

He should have looked at the territory: unusually capable people pursuing an important problem he could see firsthand.

FOLLOW EXCEPTIONAL PEOPLE

When unusually smart, driven people you deeply respect are gathering around a problem, pay attention.

A market may look unfashionable. The company may sound strange. The category may not exist yet.

The quality of the people, the problem they see, and the speed at which they learn can reveal more than public consensus.

FIND THE FRINGE

Many foundational technologies started at the edge.

Personal computers looked like a hobby for weird enthusiasts. Early social products looked trivial. New categories often begin as toys, obsessions, or niche communities.

A useful signal: if you can find the edge of the conversation but not the intellectual edge, there may still be unexplored territory.

FIND YOUR PEOPLE EARLY

Earlier generations often had to wait until college, work, or geographic relocation to find people who shared unusual interests.

Now niche communities form instantly across Reddit, X, forums, group chats, and specialized networks.

Founders can use this to find collaborators, early users, critics, and people obsessed with the same strange problem.

FOUNDERS & COMMUNITY

IDEAS + EXECUTION BEAT ACCESS

Garry credits YC’s founders with reducing the old social gatekeeping of Silicon Valley.

Instead of needing the right school, party, or introduction, founders could apply through a simple process and be judged by builders.

The principle: where you came from matters less when the system can evaluate what you think, build, and execute.

FOUNDER COMMUNITY NEEDS HONESTY

Founding is solitary.

When your best customer leaves, your strongest engineer quits, or your co-founder loses confidence, generic networking is not enough.

You need peers with whom you can be completely real.

A high-trust founder community is operational infrastructure because it gives you context, emotional stability, and better decisions.

WHY CO-FOUNDERS HISTORICALLY MATTERED

YC historically preferred co-founders because another committed person is evidence that the idea can attract belief.

One person dancing alone can look irrational. The second person changes the signal.

Co-founders also divide uncertainty, provide complementary judgment, and help sustain conviction during periods when the company looks impossible.

AI CHANGES THE SOLO-FOUNDER EQUATION

Garry argues that vibe coding and agentic coding materially change founder leverage.

A single capable person can now produce far more software, research, testing, analysis, and operations than before.

This does not eliminate the value of co-founders. It expands what a founder can attempt before building a large human organization.

THE AI-NATIVE FOUNDER

DO NOT COPY YESTERDAY’S SUCCESS

A dangerous instinct is to imitate the business model that worked for the previous generation of founders.

AI changes cost structures, team size, product speed, and competitive moats.

Garry’s warning: do not chase someone else’s past success. Build for the rules forming now, because the assumptions behind older successes may already be decaying.

PURE PER-SEAT SAAS IS LESS DEFENSIBLE

Garry questions whether a pure per-seat SaaS model will remain as durable over the next 5–10 years.

His implication:
1. SaaS can still be an entry point.
2. The real moat should grow elsewhere.
3. Consider proprietary data, network effects, workflow depth, distribution, or another compounding advantage.

CODE IS NO LONGER PRECIOUS

Previously, shipping software required specs, engineers, coordination, QA, testing, and significant time.

With coding agents, much of that work can be compressed.

This changes founder behavior. You can test small ideas, build prototypes, create internal tools, and discard weak attempts without treating every line of code as a scarce asset.

BUILD TRIVIAL THINGS ON PURPOSE

You do not need every AI experiment to become a company.

Build small, low-stakes things simply to understand what the models can do.

Use different models. Create tools. Automate tiny workflows. Test agents.

The purpose is to develop intuition. Founders who use the technology deeply will see product possibilities before people who only read about it.

AGENCY + TASTE BECOME MORE VALUABLE

If software becomes cheap to create, technical production alone becomes less differentiating.

The scarce skills shift toward:
1. Choosing worthwhile problems.
2. Knowing what “good” looks like.
3. Acting without permission.
4. Iterating rapidly.
5. Rejecting mediocre output.

Agency and taste improve through repeated practice.

YOU CAN BECOME MORE HIGH-AGENCY

Garry describes himself as a late bloomer who made major career mistakes and learned from them.

His point: agency is not necessarily fixed at birth.

You can inspect a poor decision, identify the belief that caused it, and change your future behavior.

A founder’s decision system can improve when mistakes become explicit lessons instead of hidden regrets.

BUILD BUSINESS LOOPS

THE MOST IMPORTANT LOOPS ARE BUSINESS LOOPS

Coding loops matter. Business loops matter more.

The goal is not simply to use AI for one task. The output of the task should improve how the company performs that task next time.

Example:
Do research → inspect errors → improve instructions → save the method → rerun automatically → learn from the next result.

TOKEN-MAX THE FOUNDER

Garry argues that founders and CEOs may rationally spend much more on high-context agents than ordinary users.

The reason is leverage: better decisions at the top affect product, hiring, strategy, operations, and capital allocation.

His framing is provocative: use enough compute and context to experience capabilities that may become normal only years later.

SKILLIFY EVERY REPEATED WIN

When you solve a difficult task, do not leave the solution inside one chat or one employee’s memory.

Turn it into a reusable asset:
1. Instructions.
2. Markdown or structured knowledge.
3. Code where needed.
4. Tests.
5. Automation or a scheduled run.

The result becomes a repeatable company capability.

A MARKDOWN FILE CAN BEHAVE LIKE AN EMPLOYEE

Garry’s shorthand: “a markdown file is an employee.”

The idea is not that a file literally replaces a person. It captures the method an agent needs to perform a job repeatedly.

A strong skill file can preserve context, rules, examples, checks, and expected outputs so the process becomes more consistent every time it runs.

DO THE PROCESS ONCE — PERFECTLY

Start with one business process: sales research, QA, customer support, reporting, recruiting, content, finance, or analysis.

Perform it with the agent. Correct every mistake. Improve the instructions. Add checks.

Once the result is reliable, preserve the workflow so the company does not restart from zero every time.

EVERY FUTURE ERROR BECOMES A BUG FIX

A reusable agent workflow improves differently from a human memory-based process.

When a future case fails:
1. Identify the failure.
2. Fix the instruction, logic, data, or test.
3. Save the correction.
4. Rerun.

If the improvement is retained, the organization becomes less likely to make the same exact mistake again.

AUTOMATE THE BOTTLENECK YOU CAN SEE

Garry describes building systems by repeatedly asking: where am I still doing manual verification, coordination, or decision work?

Each remaining bottleneck becomes the next automation target.

The pattern is recursive:
automate → observe new bottleneck → create tool/instruction → test → automate again.

TINY TEAMS CAN ATTEMPT HUGE OUTCOMES

Garry cites companies reaching very high revenue with only two or three people supported by hundreds of reusable agent skills.

The exact outcome will vary by company. The structural lesson is larger: headcount is becoming less tightly coupled to output.

A small team that systematizes work can compete at a scale that previously required departments.

AI-NATIVE MANAGEMENT

SCALE CREATES A CONTEXT PROBLEM

One reason companies slow down is simple: the organization becomes too complex for any one person to understand continuously.

Meetings multiply. Information fragments. Teams form local interpretations. Dependencies become invisible.

AI memory and retrieval can help reconstruct context across a company and surface what leadership is missing.

PROVENANCE MATTERS

As agents accumulate knowledge, conflicts appear.

Two documents may disagree. An old fact may contradict a newer one. A summary may be detached from its source.

Garry highlights provenance: know where information came from, when it was created, and which source should win.

AI-native companies need maintenance loops for their organizational memory.

EXPERIENCE MAY COMPOUND MORE IN THE AI ERA

Garry sees a possible advantage for founders in their 30s, 40s, and beyond.

Someone who has already built teams, products, and companies knows what failure patterns look like.

Give that person strong agents and the experience can be multiplied across research, execution, and management.

AI can amplify accumulated operating judgment.

USE AI TO SURFACE PRODUCTIVE CONFLICT

Healthy organizations need disagreement. Human conflict often adds status, fear, politics, and emotion.

AI can help compare competing approaches, summarize evidence, identify contradictions, and surface unresolved decisions.

The goal is not to remove human judgment. It is to make disagreement more legible so teams can resolve it using better context.

LEADERSHIP WITH DEEP CONTEXT

Garry describes a CEO using agents to review meeting transcripts across direct reports and deeper layers of the company.

That can reveal blockers, conflicts, and decisions without attending every meeting.

A leader can enter a critical discussion with weeks of context, make a focused decision, and leave the team to execute.

MANAGEMENT BECOMES RETRIEVAL + JUDGMENT

The AI-native executive does not need to personally sit inside every information flow.

A better system:
1. Capture meetings and operational data.
2. Build searchable memory.
3. Retrieve the right context for each decision.
4. Compare claims with ground truth.
5. Intervene only where judgment is actually needed.

HUMAN MEMORY IS A STRUCTURAL BOTTLENECK

Traditional companies assume humans can only hold a small number of active concerns at once.

That limitation shapes meetings, hierarchy, middle management, reporting, and bureaucracy.

Agents can retain and retrieve far more context. This creates a chance to redesign the company itself instead of merely adding AI to old workflows.

MAKE PRODUCTS 10×, 100×, 1000× BETTER

Garry’s aspiration is not merely to make existing work cheaper.

AI should create dramatically better products and services: more responsive, more personalized, more informed, and more capable.

The founder question is not only “What can I automate?”

It is also: “What becomes possible when intelligence and execution are far more abundant?”

REDESIGN THE COMPANY

BUREAUCRACY WASTES HUMAN POTENTIAL

Garry recalls needing extraordinary effort at Microsoft simply to get another team to address a blocking bug.

The problem was not lack of intelligence. It was layers, fiefdoms, priorities, and missing context.

Large organizations can turn simple coordination into weeks of friction.

Startups should design systems that prevent this from forming.

A STARTUP CAN — SO IT MUST

A large incumbent may struggle to redesign itself around agents, loops, and continuously available context.

A startup begins with fewer constraints.

That creates an obligation: organize for speed from day one.

Build shared memory, automate coordination, expose blockers, shorten feedback loops, and let a tiny team behave like a much larger one.

AGENTS CAN ABSORB COORDINATION WORK

One possibility discussed in the conversation is shifting more organizational work to agents:

• Track dependencies.
• Summarize status.
• Surface conflicts.
• Maintain context.
• Route issues.
• Check completion.

Executives set direction. Individuals execute. Agents can increasingly help coordinate the layer between them.

AI CAN ERASE THE API LINE

Traditional systems often force people to obey rigid software workflows.

Garry imagines a more conversational model: an agent encounters a bug, reports it, receives a workaround, and continues.

This turns software from a one-way command structure into a feedback loop.

Users and agents can influence the tools they depend on while work is happening.

TOYOTA’S LESSON: LET THE WORKER IMPROVE THE LINE

Garry connects AI-native work to the Toyota Production System.

The person closest to the work often has the strongest context for improving the process.

A powerful system lets that knowledge change the workflow.

With agents, each correction can become reusable organizational knowledge instead of disappearing after the individual task is finished.

CHAPTER 7 — HOW FAST WILL THIS CHANGE?

HUMANS ARE THE ADOPTION BOTTLENECK

Garry argues that the limiting factor may increasingly be human organizations rather than raw model intelligence.

Companies, governments, institutions, regulations, incentives, and habits move slowly.

This is important for founders: the technical future can arrive before customers and institutions are ready to reorganize around it.

THE 20-YEAR ADOPTION WINDOW

Garry compares today’s AI-native young builders with earlier generations that grew up native to the web and mobile.

Those generations eventually came to run major institutions.

His optimistic view: broad transformation may take decades, not months.

That slower transition creates time for people, companies, and society to adapt.

INCUMBENTS STILL HAVE REAL MOATS

AI increases startup leverage, but large institutions do not disappear automatically.

Incumbents have customers, distribution, regulation, capital, data, contracts, habits, and infrastructure.

A startup should respect those structural moats while exploiting the areas where a smaller organization can move faster, learn faster, and serve users better.

THE NEXT COMPUTER

THE INTERFACE MOVES TOWARD VOICE + MEMORY

Garry expects the near-term computer to look familiar, but he finds it hard to believe the current form factor is permanent.

He points toward voice, memory, computer use, ingestion, and richer personal context.

The emerging assistant is not only something you prompt. It increasingly understands what you are doing and what matters to you.

THE REAL PRODUCT IS CONTEXT

A powerful personal AI would understand your goals, fears, preferences, history, work, and current situation.

It could continuously search for ways to help rather than waiting for isolated prompts.

That requires memory, permissions, trust, retrieval, privacy, and reliable context management.

The race is not only for smarter models. It is for better harnesses.

2027: THE “HARNESS WARS” PREDICTION

Garry predicts that 2027 may become a period of intense competition around AI harnesses: the systems that connect models to memory, tools, context, computer use, workflows, and users.

This is a prediction from the conversation, not a certainty.

Founder opportunity may increasingly sit in how intelligence is packaged and directed.

CONSUMER AI NEEDS BETTER ECONOMICS

Many consumer software products rely on free trials or very low-cost distribution.

High model costs make that difficult.

As capable models become cheaper, founders can build more ambitious consumer AI products with persistent intelligence, richer interaction, and broader access.

Falling inference cost can unlock categories that are currently uneconomic.

BUILD A BETTER WORLD

ATTEMPT THE BETTER WORLD

Garry rejects the need to promise a perfect technological utopia.

The standard can be simpler: use technology earnestly to make products, services, work, and institutions better.

A founder can hold ambition without pretending every consequence is solved.

The job is to build something meaningfully better and keep correcting what fails.

WORK SHOULD BECOME LESS BUREAUCRATIC

If agents absorb reporting, coordination, retrieval, repetitive analysis, and process maintenance, people can spend more time on judgment, relationships, creativity, craft, and direct execution.

The promise is not simply “do more work.”

It is to remove the organizational friction that consumes attention without improving the product or the customer outcome.

ACT LOCAL

The final part of the conversation shifts from startups to civic life.

Garry’s principle is local action: care about the people and systems immediately around you.

Technology may increase capability, but human coordination still requires people to participate, speak up, organize, vote, and improve institutions directly.

THE NEW FOUNDER OPERATING SYSTEM

THE 10 NEW RULES

1. Trust direct experience.
2. Stop chasing what is hot.
3. Follow exceptional people.
4. Explore the fringe.
5. Build constantly with AI.
6. Increase agency and taste.
7. Turn wins into reusable skills.
8. Build business loops.
9. Redesign the organization for agents.
10. Use the leverage to build something meaningfully better.

THE FOUNDER’S WEEKLY LOOP

Every week ask:

1. What did I learn directly from users?
2. What bottleneck did I remove?
3. What process did I turn into a reusable skill?
4. What did the agent get wrong, and did I fix it permanently?
5. What important truth am I ignoring because consensus says otherwise?
6. Did the product become materially better?

FINAL IDEA

AI does not automatically create a great company.

The advantage goes to founders who change how they think, build, decide, learn, and organize.

Use AI as leverage. Preserve direct contact with reality. Capture knowledge into repeatable systems. Keep the team small enough to move fast and ambitious enough to build what old organizations cannot.

HOW TO BUILD IN THE AI AGENT ERA

AI is moving from answering prompts to completing complex work.

The opportunity is no longer just building a better model. It is designing the system around it: tools, context, memory, evaluators, hardware, and clear objectives.

THE NEW AI REALITY

AI HAS REACHED JUNIOR ENGINEER LEVEL

Jeff Dean believes modern coding agents are now capable enough to match many definitions of a junior engineer.

Their progress is also moving faster than expected. The next advantage comes from giving them longer tasks, better tools, and clearer operating rules.

AGENTS WILL WORK FOR DAYS

Most people still imagine an AI task lasting minutes or hours.

The bigger shift is agents that can work for days or weeks: rebuilding software in another language, testing alternatives, improving performance, and completing projects that once required an entire team.

THE NEXT LEAP: AI IMPROVING AI

By 2027, more ML systems may improve themselves through automated loops.

The system breaks a large problem into subproblems, runs many experiments, evaluates the results, combines the strongest solutions, and produces a better system with less human intervention.

INFERENCE IS THE NEW BOTTLENECK

THE PRODUCT CHANGES WHEN WAITING DISAPPEARS

Agent products feel limited when every step requires waiting.

Jeff Dean asks founders to imagine inference with dramatically lower latency. A system that responds 50× faster would not feel like a chatbot. It could operate continuously, interact naturally, and execute in real time.

AI PROBLEMS ARE OFTEN SYSTEMS PROBLEMS

Many teams call a limitation a model problem when the real bottleneck is hardware, memory bandwidth, data movement, or network communication.

Understanding the whole stack helps founders avoid expensive model changes when the better answer is a faster system design.

MOVING DATA COSTS MORE THAN COMPUTING

A calculation can be cheap. Moving the data required for that calculation can consume roughly 1,000× more energy.

This is why AI systems use batching. They process many tokens together to spread the data-movement cost, even though batching can reduce responsiveness.

INFERENCE NEEDS SPECIALIZED HARDWARE

Training can tolerate delay. Inference cannot.

Future hardware may minimize data movement, use extremely low-precision operations, and support fewer unnecessary features. The goal is not a machine that does everything. It is a machine that serves intelligence quickly and efficiently.

CONTEXT ENGINEERING BECOMES THE PRODUCT

THE MODEL IS ONLY ONE COMPONENT

A powerful AI product combines a model with retrieval, memory, tools, workflows, and evaluation.

The model must know what information to fetch, which tool to use, how to divide the task, when to try another approach, and how to judge whether the result is good.

CONTEXT IS CLEARER THAN TRAINING DATA

Training data is compressed into billions of parameters. The model does not recall every detail clearly.

Information placed directly in the current context is easier to use. This gives small teams a major advantage: they can improve results without training a frontier model.

TURN HUMAN METHODS INTO AI SKILLS

Jeff Dean and Sanjay Ghemawat encoded their performance-optimization process as an AI skill.

The agent learned to run benchmarks, change code, measure improvements, inspect trade-offs, and iterate. The core method was simple: give the model the workflow an expert would follow.

IMPROVE THE SYSTEM BY STUDYING FAILURES

Use agents on real problems and watch where they fail.

The fix may be a better instruction, a missing example, a new tool, a clearer evaluation rule, or a reusable skill. Context engineering improves when every failure becomes a permanent upgrade to the system.

RELIABLE AGENTS NEED STRUCTURE

AGENTS DRIFT OUTSIDE THEIR COMFORT ZONE

Agents often work well for the first few steps, then become less reliable.

As the task moves beyond patterns the model knows, performance can fall quickly. Long-running systems need guidance that keeps the agent on a clear, well-lit path.

USE MULTIPLE AGENTS TO SEARCH SOLUTIONS

A stronger system can send several agents to try different approaches.

Another agent or evaluator compares the results, keeps promising paths, and discards weak ones. This turns inference into a search process and makes complex workflows more reliable.

SKILLS KEEP AGENTS INSIDE THE WORKFLOW

A general model may not know your internal software, process, or quality standard.

Skills can teach it how to fetch logs, review code, run tests, access data, follow approval rules, and complete work in the correct order. This is how a general model becomes useful inside a specific company.

WHERE SMALL STARTUPS CAN WIN

WIN THROUGH DOMAIN DEPTH

Large AI companies build systems that work across many domains.

A small team can win by focusing on one painful workflow and designing a better interface, better data, better skills, stronger evaluation, and higher accuracy for that exact user.

LOOK FOR 0–1% MODEL SUCCESS

Test frontier models on the problem before building.

If they already succeed 20% of the time, the capability may improve quickly and absorb your product. A stronger opportunity is a task where general models currently succeed almost never.

PRIVATE DATA CREATES DEFENSIBILITY

A general model may organize public information well. It cannot automatically understand a user’s private history, company systems, customer records, or operating decisions.

Products that securely organize and activate this unique data can create durable value.

SPECIALIZED MODELS STILL MATTER

Some problems need a focused model trained for one scientific or technical domain.

AlphaFold is the pattern: a specialized system can outperform general intelligence on a narrow, valuable problem. Similar opportunities may exist in materials, chip design, biology, and other expert fields.

BECOME AN AI-NATIVE FOUNDER

CLEAR SPECS MULTIPLY AGENT OUTPUT

Agents perform better when the target is explicit.

Define the objective, inputs, constraints, expected behavior, tests, edge cases, and acceptance criteria. The clearer the specification, the less the agent must guess and the more work you can delegate safely.

EXISTING SOFTWARE IS A PERFECT SPEC

Agents are strong at translating software between programming languages because the original system is a detailed specification.

The code, behavior, and tests define the target. The agent can rebuild it, compare outputs, fix differences, and continue until the new version behaves correctly.

THE SCARCE SKILL IS TASTE

When agents can write most of the code, execution becomes abundant.

The scarce skill is choosing the right problem, defining the right product, recognizing quality, and knowing what not to build. Excellent execution on an unimportant problem still creates little value.

BUILD BETTER TASTE

RECORD WHAT YOU THINK WILL MATTER

Write down several problems or technologies you believe will become important in the next 12 months.

Later, review what actually happened. Which ideas became valuable? Which were ignored? Which assumptions were wrong? Repeated prediction and review builds practical taste.

QUESTION ASSUMPTIONS FROM FIRST PRINCIPLES

Do not stay anchored to how a problem is solved today.

Ask what would happen if speed improved 10×, cost fell 100×, errors became acceptable, or a constraint disappeared. Most thought experiments fail, but the successful ones can create new system architectures.

MAPREDUCE CAME FROM REMOVING NOISE

Google’s large-scale jobs mixed simple computation with complex code for parallelism, failures, and checkpointing.

Jeff Dean and Sanjay Ghemawat separated the repeated infrastructure from the actual task. That abstraction became MapReduce and made distributed computing easier to use.

TPU CAME FROM SIMPLE NAPKIN MATH

Better speech recognition meant people would use it more. Serving that demand on CPUs could require a dramatically larger server fleet.

That calculation exposed the bottleneck before it arrived. The response was specialized hardware that later became foundational for modern AI.

AUTOMATED SCIENCE CHANGES EVERYTHING

THE SCIENTIFIC METHOD CAN BECOME SOFTWARE

AI can automate the experimental loop: propose an idea, implement it, run the experiment, measure the result, and choose the next direction.

When thousands of experiments can run automatically, progress is measured by useful discoveries per unit of compute.

MAKE THE EVALUATOR FASTER

Some scientific simulations take hours for one answer.

Researchers can train a neural approximation using inputs and outputs from the expensive simulator. Jeff Dean described one case that became about 300,000× faster while remaining nearly as accurate.

AI WILL ORCHESTRATE SUBPROBLEMS

The most powerful systems will accept a high-level objective, divide it into subproblems, run automated experiments for each one, and combine the results.

This pattern could accelerate machine learning, science, engineering, molecule discovery, and chip design.

BUILD FOR POSITIVE IMPACT

REJECTION DOES NOT DEFINE IMPORTANCE

Jeff Dean’s work on model distillation was rejected by a conference reviewer who doubted its impact.

The idea later became central to creating smaller, faster, capable models. Good work can be early, misunderstood, or evaluated through the wrong lens.

CHOOSE WORK THAT CHANGES THE OUTCOME

Before committing years to a problem, ask one question:

If we succeed completely, will the world become meaningfully better?

A difficult project should create more than a technically impressive demo. It should unlock real capability, value, or progress.

BUILD WITH LOW-EGO PEOPLE

Choose teammates who are excellent in areas you are not, enjoy solving hard problems, and care more about the result than personal credit.

A strong team combines different tools, learns from each other, and becomes capable of building what no individual could create alone.

YOUR FOUNDER PLAYBOOK

1. Find a problem general models cannot solve.

2. Add private context, tools, skills, and evaluators.

3. Write a precise specification.

4. Run multiple agents and measurable loops.

5. Optimize the full system for speed, reliability, and impact.

THE NEXT ADVANTAGE IS SYSTEM DESIGN

The winning product will not be the one with the longest prompt.

It will combine the right problem, proprietary context, expert workflows, reliable evaluation, low-latency execution, and strong founder taste into a system users can trust.

PHYSICAL AI THAT SHIPS SAFELY

A working demo can impress people.

A real physical AI product must keep working safely through millions of unpredictable situations.

This is the playbook Waymo learned while turning autonomous driving from a prototype into a service operating at scale.

WHY PHYSICAL AI IS DIFFERENT

THE BEST AI MOMENT LOOKS LIKE NOTHING

When another driver suddenly cuts in, the best system reacts safely and smoothly.

Passengers may not even notice.

For physical AI, success is often invisible: no crash, no panic, no interruption. The task simply gets done.

MOVE FAST. SHIP SAFELY.

“Move fast and break things” works poorly when mistakes can damage property or harm people.

Physical AI needs speed with discipline.

Safety cannot be added later. It must shape the model, hardware, training, validation, and deployment from day one.

FOUR GAPS TO CROSS

Physical AI faces four structural gaps:

1. Cost of error
2. Latency
3. Data
4. Validation

Any serious robot, vehicle, or real-world agent must cross all four before it can operate safely at scale.

1. COST OF ERROR

A chatbot mistake may cost one retry.

A physical AI mistake can cost a human life.

The system must avoid dangerous actions before they happen because the real world offers no reset, rollback, or second attempt.

2. LATENCY

Digital assistants may take seconds to answer.

A car at freeway speed moves roughly 100 feet in one second.

Physical AI must sense, reason, and act in milliseconds using onboard compute that fits inside the machine.

3. DATA

Digital AI learned from the internet: a massive collection of human knowledge and labeled examples.

Physical AI has no equivalent dataset.

Teams must collect, generate, simulate, label, and continuously improve their own real-world data.

4. VALIDATION

Digital products can launch early and let users reveal edge cases.

Physical AI needs strong safety and confidence before the first public deployment.

The operating conditions must be defined clearly, tested rigorously, and expanded responsibly.

LESSON 1 : A DEMO IS ONLY 1% OF THE WORK

THE DEMO CAN FEEL LIKE THE FINISH LINE

Waymo reached major autonomous-driving demo goals in about 18 months.

It could drive day and night, through traffic, construction, highways, and city streets.

By demo standards, the problem looked solved. The product journey had barely started.

PRODUCT MEANS REPEATING IT RELIABLY

Driving ten difficult routes once is impressive.

Operating millions of miles without a human driver is a different engineering problem.

Waymo needed roughly 15 years to move from the first strong demo to a service operating at large scale.

RELIABILITY IS AN EXPONENTIAL LADDER

Reaching 90% or 99% performance is often the easy part.

Each additional nine of reliability can demand roughly ten times more effort.

The architecture, testing, redundancy, and operations must evolve at every level.

THE LONG TAIL BECOMES DAILY REALITY

An event that happens once in a million miles sounds negligible.

When the fleet drives millions of miles every week, it becomes a routine problem.

At scale, edge cases are not the edge of the product. They are the product.

COUNT YOUR NINES FIRST

A demo may need one nine.

An assistive tool may need several.

A fully autonomous public system needs many more.

Define the required reliability before choosing the architecture, budget, timeline, and launch strategy.

LESSON 2 : CHOOSE TECHNOLOGY FOR THE REQUIRED CEILING

THE FASTEST START CAN BE A TRAP

Many technologies improve quickly at first, then flatten.

Teams often choose the path with the fastest early demo and assume the curve will continue.

The danger appears when that approach reaches its ceiling below the product’s required performance.

BUILD FOR THE PRODUCT, NOT THE PROTOTYPE

A prototype can optimize for speed of learning.

A production system must optimize for the reliability ceiling it eventually needs.

Be honest about whether each technical choice supports a demo, an experiment, an assistive tool, or full autonomy.

USE COMPLEMENTARY SENSING

Cameras provide color and detail.

LiDAR measures 3D structure, even in darkness.

Radar handles weather and measures velocity.

Fusing multiple sensing modes creates a stronger view of the world than relying on one sensor alone.

REDUNDANCY MUST BE DESIGNED IN

A leaf, glare, darkness, dust, rain, or physical damage can weaken a sensor.

Physical AI needs fallback paths and redundant perception.

The goal is not only to detect failure, but to continue safely when part of the system degrades.

DESIGN FOR FUTURE HARDWARE ECONOMICS

Hardware becomes cheaper, smaller, and more capable across generations.

Do not lock the company into assumptions based only on current prices.

Design the system so sensors, compute, and vehicle platforms can be upgraded as the technology matures.

LESSON 3 : RIDE EVERY MAJOR TECHNOLOGY WAVE

ONE BREAKTHROUGH IS NOT ENOUGH

Waymo repeatedly rebuilt its driver around major advances: convolutional networks, transformers, vision-language models, and world models.

The advantage is not adopting one new technology.

It is learning how to absorb the next one without stopping deployment.

RESEARCH MUST HAVE A PRODUCTION PATH

Before launching a new technical project, define what happens if it succeeds.

How will it enter the main product?

How will it be validated?

How will it replace or simplify existing systems?

Innovation without an integration path creates expensive fragmentation.

DEMAND PERFORMANCE AND SIMPLIFICATION

A breakthrough that adds capability but creates another isolated stack can slow the company down.

The launch bar should demand two outcomes:

1. Meaningful performance gains
2. Radical simplification or unification

THE FOUNDATION MODEL BECOMES THE CORE

Waymo’s foundation model combines multimodal sensing, world understanding, action prediction, and language-aligned knowledge.

A large shared foundation moves complexity upstream, while smaller specialized models run efficiently on different vehicles and hardware.

THINK FAST. THINK SLOW. GENERATE.

The fast path handles split-second geometric danger.

The slow path reasons about deeper semantics, such as a burning vehicle or unusual scene.

The generative component predicts how others may behave and selects the vehicle’s next action.

LESSON 4 : USE STRUCTURE TO IMPROVE SCALING

GENERAL METHODS WIN WITH SCALE

The bitter lesson of AI is consistent: methods that scale with compute and data eventually outperform systems dominated by human-written rules.

High-capacity models learn richer patterns and can later be distilled into smaller models for real-time deployment.

STRUCTURE MUST CHANNEL SCALE

Structure that fights scale will lose.

Structure that channels scale can win.

The right representation should not shrink the solution space. It should make training, evaluation, and safety checks more efficient without removing the model’s ability to learn.

STRUCTURE-AUGMENTED END-TO-END

Waymo combines end-to-end learning with structured representations of the physical world.

The model still learns rich internal embeddings, while explicit structure supports validation, physics, road rules, object behavior, and measurable safety checks.

WHY THE STRUCTURE MATTERS

Structured intermediate representations enable:

1. Real-time correctness and safety checks
2. More efficient large-scale training and evaluation
3. Stronger feedback signals for metrics, loss functions, and reinforcement learning

LESSON 5 : BUILD A HIGH-FIDELITY SIMULATOR

OPEN LOOP IS NOT ENOUGH

Open-loop training asks: “In this situation, what would you do?”

Closed-loop training asks the agent to act, observe the consequence, update its world view, and act again.

Safety-critical systems must understand sequences, feedback, and counterfactual outcomes.

THE SIMULATOR IS ANOTHER AI SYSTEM

A serious simulator is not lightweight test software.

It is a world model that understands physics, traffic, weather, objects, behavior, and sensor outputs.

Its realism must be high enough to train and evaluate decisions that will later affect real people.

TRAIN ON EVENTS THAT RARELY HAPPEN

Simulation can create difficult situations that may be unsafe, expensive, or nearly impossible to collect repeatedly in reality.

A stopped car on a freeway, a plane landing ahead, an animal in traffic, or unusual weather can be generated, controlled, and tested at scale.

CLOSED LOOP CREATES CONFIDENCE

The goal is not only a realistic-looking video.

The agent must interact with a simulated world that responds to its decisions.

This allows teams to test whether one action creates a safe or dangerous sequence several steps later.

LESSON 6 : BUILD THE AGENT, SIMULATOR, AND CRITIC

YOU NEED THREE AIS

The agent acts in the real world.

The simulator creates realistic worlds and difficult scenarios.

The critic evaluates performance and explains where improvement is needed.

Together, they form the learning ecosystem behind a scalable physical AI product.

CREATE THE IMPROVEMENT FLYWHEEL

Real-world deployment generates data.

The data makes simulation more realistic.

The simulator produces harder edge cases.

The critic scores them.

The agent learns, improves, deploys again, and generates better data.

A FLYWHEEL NEEDS DIRECTION

A flywheel can spin quickly in the wrong direction.

Clear metrics determine which failures matter, which scenarios deserve more data, and which model changes create genuine improvement.

Without metrics, more data and compute can still produce confusion.

LESSON 7 : EVALUATION AND METRICS ARE THE MOAT

BUILD THE EVAL BEFORE THE PRODUCT

If the team cannot measure what “good enough” means, it is not building a product.

It is iterating on a demo.

Evaluation should define the target, expose weaknesses, prioritize data, measure progress, and determine whether the system is ready to deploy.

EVALUATE THE ENTIRE SYSTEM

Physical AI validation must cover sensors, compute, models, fallback systems, behavior, operations, maintenance, remote support, deployment rules, and real-world readiness.

Safety depends on the full system, not one benchmark score.

TRUST IS EARNED WITH EVIDENCE

Customers, communities, and regulators trust systems that repeatedly prove they work.

Models can be copied. Algorithms can leak.

Years of real-world operation, rigorous evaluation, safety data, and transparent evidence are much harder to replicate.

SAFETY BECOMES A BUSINESS ADVANTAGE

When a physical AI product demonstrates safer outcomes across hundreds of millions of autonomous miles, safety is no longer only a technical metric.

It becomes brand trust, regulatory confidence, customer adoption, operational learning, and a durable competitive advantage.

THE COMPLETE PLAYBOOK : SEVEN LESSONS THAT REINFORCE EACH OTHER

THE SYSTEM COMPOUNDS

Required nines define the technology ceiling.

The right architecture enables repeated upgrades.

Structure improves scaling and validation.

Simulation exposes rare events.

The agent, simulator, and critic create a flywheel.

Metrics keep the flywheel moving toward safety.

PHYSICAL AI IS ENTERING ITS DECADE

Digital AI transformed screens, software, and knowledge work.

The next major wave will increasingly move through vehicles, robots, factories, logistics, healthcare, and infrastructure.

The opportunity is massive, but the standard must remain: move fast and ship safely.

REMEMBER WHO YOU ARE BUILDING FOR

The goal is not the most impressive demo.

The goal is a system people can trust with real tasks, real environments, and real consequences.

Build for the customer, the mission, and the human life affected by every decision.

THE MINDSET THAT BUILT NVIDIA

NVIDIA did not begin with the right technology.

It survived because the team confronted reality, learned what it did not know, and rebuilt before time ran out.

CONFRONT REALITY

A THOUGHTFUL BET CAN STILL FAIL

NVIDIA began with a bold idea: add an accelerator to the PC and turn every computer into a game console.

The idea was carefully reasoned. The team believed in it.

The algorithm was still fundamentally wrong.

SAY IT BEFORE IT KILLS THE COMPANY

By 1995, dozens of competitors were building PC graphics products.

NVIDIA had little time left.

Jensen told the company the truth: the technology did not work, and avoiding that fact would end the company.

NOT KNOWING IS NOT THE END

The team did not know the correct approach either.

Jensen bought three textbooks on OpenGL and graphics pipelines, brought them back, and the engineers learned from them.

A company that later led computer graphics restarted from books.

THE DURABLE ADVANTAGE

Technology keeps changing.

A specific tool, chip, or algorithm can become obsolete.

The durable advantage is the ability to confront reality, learn quickly, and act before the window closes.

BUILD ON A BELIEF

THE ORIGINAL INSIGHT WAS BIGGER

NVIDIA’s first graphics method was wrong, but its deeper belief was right.

General-purpose CPUs could be augmented with accelerators to solve problems that were otherwise too difficult.

DO NOT BUILD ONLY A CHIP

The company learned that success was not about producing a great chip alone.

It had to understand the algorithm, software, tools, applications, and complete system required to make acceleration useful.

GREAT COMPANIES SEE DIFFERENTLY

A strong company begins with a perspective about an important future that few others fully see.

The belief must be deep enough to survive doubt and difficult enough that execution creates real advantage.

ALEXNET WAS MORE THAN ALEXNET

When AlexNet appeared, Jensen did not see only an image-recognition result.

He saw deep learning as a universal function approximator and a new way to build software.

That changed the processor, middleware, applications, and industries around it.

ASK: IF THIS, THEN WHAT?

A breakthrough becomes valuable when you follow its consequences.

If this capability improves, what becomes possible?

Which industries change?

Which bottlenecks move?

Which new system must now be built?

FOUNDER MODE

CURIOSITY IS THE STARTING POINT

Jensen begins with questions.

When nearby answers are not satisfying, he goes directly to papers, researchers, engineers, and first principles.

The goal is not control. It is understanding.

THE CEO SERVES THE COMPANY

A CEO should learn enough to give the organization useful insight.

The job is to simplify a complex shift, explain why it matters, and help teams turn it into action.

YOU MUST FEEL THE TECHNOLOGY

Fast-moving technology looks chaotic from far away.

When you work close to the details, patterns become readable.

Like a surfer, a founder learns the wave by entering it, not by studying reports from shore.

BUILD THE CAR YOU CAN RACE

A founder is building an F1 car they must personally drive.

The company’s processes, communication, and structure should amplify the founder’s strengths and decision speed.

The next CEO can reshape it later.

KEEP TWEAKING THE MACHINE

Founder mode is not one fixed management style.

It is the continuous redesign of processes, information flow, and decision paths so the company can move faster without losing understanding.

SYSTEMS THINKING

LOW-LEVEL WORK WILL BE AUTOMATED

Many implementation tasks will be performed agentically.

The valuable human skill becomes defining the problem, constraints, inputs, outputs, information flow, bottlenecks, and architecture of the complete system.

THINK IN CONSTRAINTS

A systems thinker asks:

Where does information enter?

What limits throughput?

Is the constraint compute, memory, networking, data, coordination, or control?

The right answer determines the architecture.

AGENTS NEED FINE-GRAINED CONTROL

Agents do not need to be perfect before they become useful.

They need controllability.

A human should be able to change one instruction, component, pixel, layer, or connection and produce a specific delta without rebuilding everything blindly.

USE AI TO LEARN AI

NVIDIA lets teams use different coding and agent tools.

The goal is not to force one winner too early.

Broad usage helps the company move faster and reveals how future systems should be designed.

EVERY COMPANY CAN BUILD ITS OWN AI

Cloud AI services remain useful, but companies can also build domain-specific agents around their own knowledge, tools, workflows, and standards.

The strongest advantage may come from AI shaped around how the company actually operates.

AI CHANGES THE WORK

AI AUTOMATES TASKS

A job is not one task.

It is a purpose made of many tasks.

AI can remove repetitive cognitive work while people continue handling judgment, coordination, responsibility, relationships, and new problems.

PRODUCTIVITY CAN CREATE DEMAND

When coding, medical analysis, or legal processing becomes faster, organizations can serve more ideas, patients, and cases.

Higher productivity can expand output, which creates demand for more people across the wider system.

PHYSICAL AI IS THE NEXT PLATFORM

If AI can generate believable motion, it can begin learning how robots should move.

The challenge is grounding that motion in causality, friction, tension, physics, simulation, reinforcement learning, and real electromechanical systems.

START WHERE ECONOMICS ARE REAL

NVIDIA viewed autonomous vehicles as an early robotics market with large demand, standardized technology, and real economic value.

A platform grows faster when its first use case can fund the learning flywheel.

WHAT TO LEARN NOW

LEARN THE HARD SCIENCES

Coding mechanics may become increasingly automated.

Physics, chemistry, biology, computer science, engineering, and their intersections remain essential because they help people define and solve harder problems.

STUDY MARKETS AND SOCIETY

The next opportunities sit where technology meets unmet demand, social constraints, industry gaps, and human behavior.

Technical capability matters most when connected to a meaningful problem.

ORCHESTRATE MILLIONS OF AGENTS

The future builder may manage systems of agents instead of manually completing every task.

The key skill is translating a large mission into architecture, standards, feedback loops, evaluation, and coordinated execution.

ENTREPRENEURIAL RESILIENCE

YOU WILL NEVER KNOW EVERYTHING

Jensen feared investor questions because he did not know every answer.

That uncertainty never fully disappears.

The founder’s advantage is not complete knowledge. It is confidence in the ability to learn.

HOW HARD CAN IT BE?

Do not imagine every future difficulty at once.

That turns uncertainty into anxiety and delay.

Begin with: “How hard can it be?”

Then let the real difficulty arrive one solvable piece at a time.

WIN TODAY

You do not need to overcome the entire journey today.

You need to get through this morning, solve today’s problem, and keep moving toward tomorrow.

Consistency converts fear into progress.

LEARNING IS THE SUPERPOWER

Believe in something important.

Start moving.

Learn what the next step requires.

Use agents, books, experts, experiments, and first principles.

Stay with the mission long enough for compounding to become visible.

NEVER A BETTER TIME TO START

AI is changing what a small team can build.

Work that once took months can now be compressed into minutes. The opportunity is not to build smaller companies. It is to attempt bigger problems with fewer people, faster cycles, and far more leverage.

THE NEW STARTUP WINDOW

Great startups often appear when technology shifts quickly, costs fall, and incumbents lose their advantage.

The internet, app stores, and mobile created earlier waves. AI is creating another major opening where expertise, production, and experimentation are dramatically cheaper.

WHY NOW

STARTUPS ARE MORE POWERFUL

Twenty years ago, a YC startup could spend three months building one product.

Today, a coding agent may reproduce that work in minutes. Founders can now coordinate software, research, design, operations, and analysis at a scale previously reserved for large companies.

AI RAISES THE AMBITION BAR

A weak response to AI is asking which old startup ideas are now obsolete.

A stronger response is asking what company was impossible to build before. AI gives founders capabilities that once required large teams. The advantage belongs to people willing to attempt harder and more valuable problems.

GOLDEN AGE OF STARTUPS

Hard technology remains difficult, but AI reduces barriers around research, prototyping, hiring, and coordination.

A project that once required a large institution may now begin with a small team using agents and compute. This creates room for a new generation of deeply ambitious companies.

THE STARTUP IS NOT OVER

Some people fear powerful models will absorb every valuable company.

The opportunity may be larger. New startups can become more valuable and impactful because they can build faster, reach further, and solve problems that were previously uneconomic. Intelligence on demand expands the design space.

AI LEVERAGE

BUILD WITH AGENTS

A small team can now automate major parts of a startup using AI agents.

This does not remove the need for founders. It changes the founder’s job. The scarce skills become judgment, taste, agency, customer understanding, product direction, and knowing where durable value can be created.

TOOL FLUENCY COMPOUNDS

Years of experience still matter, but AI fluency is becoming a major advantage.

People who grew up using these tools can test ideas, automate operations, and learn unfamiliar fields quickly. In a fast-moving market, working effectively with new tools may outperform traditional credentials.

CYCLE TIME IS STRATEGY

When costs fall and cycle times shrink, founders can test more assumptions with less capital.

The advantage is not merely shipping faster. Faster cycles create more customer feedback, more data, and more opportunities to correct direction. Speed becomes a learning system.

DO THREE MONTHS OF WORK

If AI compresses three months of work into minutes, the goal is not to stop after one prompt.

Use the remaining time to explore more directions, improve quality, talk to customers, and build stronger systems. The productivity gain should raise the standard of what a founder attempts.

CONVICTION

FIND THE UNPOPULAR TRUTH

Many important companies begin with a belief that experts dismiss.

Develop reasonable conviction in something the market has not understood. When new evidence continues to support the idea, being misunderstood can become an advantage because competition remains limited.

OPENAI’S EARLY SECRET

For years, only a small number of people believed AGI was possible.

That disbelief gave OpenAI time to research, recruit, and build without overwhelming competition. A startup does not need everyone to agree. It needs enough strong people who understand the same emerging truth.

STUDY NEW EXPONENTIALS

People struggle to understand exponential change.

Founders should search for technologies, costs, behaviors, or capabilities improving faster than the market expects. Opportunity appears when reality has moved, but conventional wisdom still uses an older model of what is possible.

VISION CAN BE CLEAR FIRST

A very ambitious startup may begin with a clear destination and unclear first steps.

OpenAI wanted to build AGI long before it knew it would become a product company. Uncertainty is acceptable, but progress is required. Take imperfect steps that generate evidence and reveal the next move.

CONVICTION NEEDS EVIDENCE

Strong conviction should grow through data, experiments, technical progress, or customer response.

When nobody credible shares the belief and no evidence improves over time, pay attention. The goal is not blind confidence. It is a belief that survives contact with reality and becomes more precise.

PEOPLE AND NETWORKS

FIND YOUR PEOPLE

Startups are difficult and lonely to build alone.

You do not need universal support, but you need a few people who understand the mission. Strong co-founders, early teammates, and trusted peers provide complementary skills, resilience, and shared conviction when progress is uncertain.

NETWORKS COMPOUND SLOWLY

Important startup relationships often begin years before their purpose becomes clear.

Sam Altman met Greg Brockman while helping Stripe recruit him. Eight years later, they started OpenAI together. Useful networks are built through repeated trust, not last-minute transactions.

BE MILDLY HELPFUL

One of the best ways to build a valuable network is to help capable people in small, genuine ways.

You gain exposure to interesting problems, earn trust, and create future possibilities. Help should be useful now, even when there is no visible return. The long-term effects are unpredictable.

CHOOSE HIGH-DENSITY ROOMS

Founders benefit from environments where ambitious builders repeatedly meet each other.

Accelerators, technical communities, startup cities, and strong networks increase the probability of finding co-founders, early hires, investors, and collaborators. Proximity creates more chances for luck to compound.

BUILDER MINDSET

OPTIMISM CREATES MOMENTUM

Paul Graham helped early founders leave each weekly dinner believing their companies could matter.

That optimism was not empty motivation. It redirected energy toward action. In the early stage, founders often need someone who can convert confusion and rejection into a clear next move.

THE FLIGHT INSTRUCTOR MODEL

Startup advice works better like flight instruction than a classroom lecture.

A useful mentor sits beside the founder and points out the immediate mistake, missed signal, or next action. Specific feedback changes behavior faster than abstract theory because it connects directly to the current situation.

EARNESTNESS BEATS SARCASM

It is easy to mock people attempting difficult things and receive attention for a clever comment.

The cost is subtle. Sarcasm trains the mind to observe instead of participate. Put energy into building, helping, and learning. Over time, real output compounds while internet points disappear.

EXPECT RESISTANCE

The more a company threatens the existing order, the more criticism it may attract.

Dismissal does not prove an idea is correct, but it is often part of important work. Founders must hear useful criticism without allowing every attack to redirect the mission.

POWER AND RESPONSIBILITY

STARTUPS DISTRIBUTE POWER

AI can distribute capability across millions of people, or concentrate power inside a few companies and models.

Successful startups help widen access by creating alternatives, new applications, and independent economic value. Building a strong company can contribute to a healthier ecosystem.

SAFETY IS REAL

As AI systems become more capable, alignment, security, and loss-of-control risks become practical concerns.

Founders should not treat safety as someone else’s problem. Products need secure infrastructure, clear safeguards, responsible deployment, and human values guiding the systems.

PROTECT HUMAN AGENCY

A future with material abundance but no privacy, freedom, or meaningful choice would still be a failure.

Progress should increase human agency. People should gain more control over their time, work, creativity, and decisions. Technology should expand participation, not create a comfortable surveillance system.

EXECUTION

TAKE THE FIRST IMPERFECT STEP

A bold vision is useless if it never produces forward motion.

The first product, experiment, or structure may be wrong. It still matters because action generates evidence. Build something small enough to test, but meaningful enough to reveal whether the larger direction has real potential.

AMBITION NEEDS AGENCY

AI gives founders more intelligence, tools, and speed, but the founder must still choose the problem and drive the system.

Agency means noticing what matters, deciding without perfect information, coordinating resources, and continuing through uncertainty. Tools amplify the person who acts.

FAIL WITHOUT ENDING THE STORY

Early startups can fail without destroying a founder’s future.

Technology ecosystems often reward learning, persistence, and the ability to try again. Keep the ambition, but do not make every setback a final judgment. Experience, relationships, and judgment can compound even when one company fails.

ENJOY THE BUILD

Startups are stressful because the outcome is uncertain and the work feels personal.

Keep the drive, but allow room for happiness during the process. Careers unfold through many attempts, mistakes, and unexpected connections. The current company matters, but it is part of a much longer journey.

THE FOUNDER’S PLAYBOOK

1. FIND THE SHIFT

Identify the capability, cost curve, or behavior that changed.

Do not begin with a fashionable feature. Begin with the structural shift. Ask what can now be built, delivered, or distributed that was impossible or uneconomic only a few years ago.

2. FORM A STRONG BELIEF

Choose a large problem and develop a clear point of view.

Look for evidence that the market has not fully absorbed. Speak with experts, build prototypes, and track progress. Keep refining the belief until it becomes specific enough to guide product decisions.

3. GATHER THE MISFITS

Find a small group with complementary skills and shared conviction.

The team does not need to look conventional. It needs trust, speed, technical or market strength, and the willingness to continue when outside validation is limited.

4. USE AI AS LEVERAGE

Use agents to compress research, coding, design, operations, and analysis.

Then reinvest the saved time into higher standards, more experiments, and deeper customer understanding. AI should increase the ambition and quality of the company, not merely reduce effort.

5. BUILD, LEARN, REPEAT

Ship an imperfect first version, measure the response, and improve quickly.

Each cycle should answer a real question: Does the technology work? Does the customer care? Can the system scale? Evidence turns a bold idea into an executable company.

THE OPPORTUNITY

There may never have been a better time to start an ambitious company.

AI has lowered the cost of capability, shortened the path from idea to product, and expanded what small teams can attempt. Choose a meaningful problem, take the first step, and begin generating evidence.