Ambitious AI Startup Playbook

Turn a 39-minute founder interview into an AI coach that can teach, challenge, and pressure-test your startup decisions.

English | 简体中文

License: MIT Source-grounded Agent Skill

Most founder content gives you ideas to agree with.

This repository gives your AI agent a system to work through those ideas with you.

It transforms Sam Altman's 2026 Startup School interview, Never a Better Time to Do a Startup, into a reusable, evidence-grounded Agent Skill. Instead of summarizing the video, the skill can teach the principles, quiz you, stress-test a real startup thesis, and help you turn an ambitious vision into the next data-generating move.

What can it do?

The skill automatically works in four modes:

Mode What your agent does
Learn Teaches one principle at a time, checks understanding, and adapts the next lesson
Practice Gives a founder-thesis exercise, scores the attempt, and reveals hints progressively
Apply Reviews a real startup idea using observable evidence, falsifiable assumptions, and concrete next steps
Reference Finds the exact decision rule, caveat, source claim, and video timestamp

It helps founders reason about five hard questions:

  1. Leverage: If AI makes implementation radically cheaper, are you merely shipping the old idea faster—or attempting something newly possible?
  2. Timing: What changed in technology, cost, iteration speed, or incumbent advantage to make this startup possible now?
  3. Conviction: Is your non-consensus belief getting stronger because of evidence, or only because people disagree with you?
  4. Momentum: What imperfect next step will generate information even if it fails?
  5. Power: Can the product preserve a meaningful safety floor without concentrating control or reducing human agency?

Try asking

Use ambitious-ai-startup-playbook to stress-test my startup thesis.
Do not flatter me. Find the assumption most likely to break.

Teach me how to distinguish non-consensus conviction from founder delusion.
Quiz me before showing the answer.

Review my AI product for safety, concentration of power, and user agency.
Label every conclusion as source claim, project fact, or inference.

My mission is clear, but I do not know the first step.
Help me choose an action that produces useful data within one week.

Why this is more than a video summary

39-minute interview
        ↓
timestamped claims + uncertainty boundaries
        ↓
lessons + exercises + scoring rubric
        ↓
startup review playbooks + decision rules
        ↓
an AI agent that can teach and apply the material

Every consequential claim maps back to a timestamped evidence window. The package preserves uncertainty instead of silently turning a speaker's forecast, historical recollection, or normative opinion into fact.

That means the agent is instructed to:

  • separate the speaker's claims from independently verified facts;
  • avoid treating rhetorical productivity numbers as benchmarks;
  • ask what evidence would strengthen or weaken a founder's belief;
  • label adaptations to your startup as inferences;
  • keep safety, distributed power, and human agency in the same review.

See the source map for human-readable citations and the provenance ledger for claim-level evidence mappings.

Install

Clone the repository into your agent skills directory:

git clone https://github.com/Lum1104/ambitious-ai-startup-playbook.git \
  ~/.agents/skills/ambitious-ai-startup-playbook

Then start a new agent session and mention the skill by name:

Use ambitious-ai-startup-playbook to review my startup idea.

You can also clone it anywhere and ask a compatible coding agent to read SKILL.md before working with the material.

What is inside?

.
├── SKILL.md                     # Agent behavior, routing, and content index
├── chapters/                    # Four adaptive learning modules
├── exercises/                   # Founder Thesis Stress Test
├── solutions/                   # Hidden-until-attempted rubric and example
├── playbooks/                   # Startup and AI governance review workflows
├── reference/                   # Compact decision rules and caveats
├── sources.md                   # Human-readable timestamp map
└── provenance.json              # Machine-readable claim-to-evidence ledger

The instructional materials are written in English; a Chinese project overview is available in README.zh-CN.md.

A note on scope

This is an independent educational project derived from one public interview. It is not affiliated with Sam Altman, OpenAI, Y Combinator, or Startup School.

The repository contains transformed teaching material, not a transcript or video archive. It does not independently validate the interview's forecasts, historical examples, or incident descriptions, and it is not investment, legal, or safety-compliance advice.

Contributing

Corrections to timestamps, evidence boundaries, translations, exercises, and playbooks are welcome. Please keep new claims traceable, label inference clearly, and avoid adding raw copyrighted source material.

License

The original material in this repository is available under the MIT License. Third-party source material remains the property of its respective owners.

Generated by video-to-skill.