Founder Mode

Take a product opportunity from first principles through a scoped build, independent validation, launch, and learning.

Don't just read it. Run it.

Start the work here, keep the important context together, and hand it off to people or AI.

Use this Playbook

Steps

  1. 1

    Clarify the Opportunity

    ID:ZXC8

    Start with the customer problem, not a solution. State the target user, their context, the current alternative, why the problem matters now, and the desired outcome. Separate facts from assumptions; do not proceed until the problem statement and success hypothesis can be tested.

    skill: github:garrytan/gstack @ office-hours
  2. 2

    CEO Review

    ID:VM4N

    Decide whether this deserves investment now. Challenge the expected impact, strategic fit, differentiation, opportunity cost, and downside; explicitly identify what will not be built. Choose proceed, stop, or learn more. Proceed only when the scope, non-goals, and decision owner are clear.

    skill: github:garrytan/gstack @ plan-ceo-review
  3. 3

    Engineering Review

    ID:WK3N

    Translate the approved scope into an implementation approach. Examine the architecture, data model, interfaces, dependencies, migrations, failure modes, security implications, and test strategy. Record material trade-offs and risks. Do not begin the build with an unresolved technical blocker.

    skill: github:garrytan/gstack @ plan-eng-review
  4. 4

    Build

    ID:19KK

    Implement only the approved scope and follow the engineering plan. Keep behavior changes covered by tests and record deviations, discoveries, or new requests for a later decision instead of silently expanding scope. Finish with a working change, its tests, and a concise summary of modified files.

    agent: github:openai/codex @ coding-agent
  5. 5

    Independent Review

    ID:66IX

    Review the change independently from its implementation. Check correctness, security, maintainability, test coverage, operational impact, and alignment with the approved scope. Classify findings by severity, return blockers with actionable fixes, and verify that every blocker is resolved.

    skill: github:garrytan/gstack @ review
  6. 6

    Product QA

    ID:LTUW

    Validate the real user journey in the browser or relevant client. Exercise the primary path, important permissions and error states, and the edge cases identified in planning. Capture reproducible evidence for failures. Pass QA only when the intended experience works without a release-blocking defect.

    skill: github:garrytan/gstack @ qa
  7. 7

    Ship

    ID:7K7J

    Confirm the latest tests, review, and QA evidence before preparing the release. Summarize user-facing changes, deployment steps, rollback considerations, and ownership after launch. Do not ship with an unresolved blocker or an unknown deployment prerequisite.

    skill: github:garrytan/gstack @ ship
  8. 8

    Retro

    ID:MFYZ

    Review the outcome against the original hypothesis. Record what worked, what was surprising, what created delay or risk, and the smallest changes that will improve the next cycle. Convert durable learnings into owned follow-up actions.

    skill: github:garrytan/gstack @ retro

Used in Public Cases

Published work that started from this Playbook.