Give AI bounded jobs.
AI is strongest when the task has context, constraints, expected output, and review criteria.
Back to Playbook
Notes & essays
AI is leverage, not authorship
How I use AI collaborators while keeping human judgment explicit.
01Context
AI helps; humans own.
AI can compress the distance between thought and artifact. That is the leverage. But it does not remove authorship, judgment, or responsibility. In practice, it makes those things more important because more can be produced before anyone has checked whether it is right.
My approach is to make the human-owned parts explicit: the problem, the taste, the constraints, the review criteria, the risk boundary, the verification, and the final claim. AI helps produce drafts, code, tests, alternatives, and synthesis. The product judgment stays accountable.
02Model
Leverage with explicit ownership.
The model is the part of the article where the idea becomes usable: a sequence of decisions, artifacts, or checks that can guide real product work.
Treat output as a draft until tested.
Generated code, copy, strategy, and analysis need review against source truth and product intent.
Keep final claims human.
The human owns what is shipped, what is claimed, what risk is accepted, and what remains uncertain.
The model moves from question to artifact to evidence.
- DelegateGive AI bounded jobs.
- VerifyTreat output as a draft until tested.
- OwnKeep final claims human.
03Core distinction
Leverage increases responsibility.
The faster I can move from idea to artifact, the more important it becomes to know which artifact matters, what quality bar applies, and what evidence is enough.
Across the projects, the strongest moments are not only the things that shipped. They are the decisions to analyze before implementing, revert a plausible but wrong migration, pause a risky commerce expansion, and turn repeated issues into shared rules.
04Practical rule
AI should make product judgment more visible, not less.
The best AI-assisted workflow leaves behind briefs, acceptance criteria, review notes, validation checks, and launch decisions. Those artifacts make the work inspectable.
The public version of this is claim discipline. It is stronger to say that I led framing, prioritization, experience design, acceptance criteria, release decisions, and AI-assisted execution than to imply the model independently built or validated the product.
05Artifacts
What this leaves behind.
I use artifacts as evidence of thinking. They make product judgment reviewable, reusable, and easier to connect back to the work.
Product brief
Turns a vague idea into users, scope, acceptance criteria, constraints, and decision rights.
Real-codebase prototype
Proves the riskiest workflow in actual implementation context instead of relying only on static mockups.
Launch gate
Collects security, privacy, payment, content, support, and monitoring checks before public exposure.
Learning loop
Connects user signal, analytics, validation scripts, release notes, and roadmap decisions.