Table of Contents

Vibe coding lowers the cost of generation. It does not lower the cost of responsibility.

AI does not remove the need for careful engineering. It makes careful engineering cheaper.

AI gives you more attempts. It does not tell you which attempt deserves to become real.

The model is allowed to be creative before the gate. It is not allowed to be authoritative after the gate.

Hallucination is inevitable. Silent commit is optional.

Mixed output must be routed. Treating the whole answer as equally trustworthy is the original sin of vibe coding.

A better prompt is not a substitute for a safer system.

Remove the hazard before teaching the agent to tiptoe around it.

Do not give the agent a chainsaw when a socket wrench will do.

A reflex is a validation step that fires before the system has to think.

Process is not the safety system. Process is how humans remember to use the safety system.

Prompt discipline improves behavior. It does not enforce behavior.

Intent becomes real only after it crosses the commit boundary.

If the next agent needs to know it, put it in the repo.

Every bug fixed without a test is a lesson the next agent can forget.

A schema is a boundary the model cannot argue with.

Every AI session that modifies code should run on a branch, not main.

The first commit should be the safety system, not the feature.

Do not ask the agent to build a feature until the feature has a shape.

Never let the agent fix a bug it cannot reproduce.

The safer the test harness, the more aggressive the refactor can be.

Docs that lie are worse than no docs, because they replace the silence that forces investigation with the confidence that prevents it.

Do not deploy what you cannot roll back.

The model may explain the math. The math engine owns the answer.

The model narrates the world. The engine decides what is true.

Mixed output becomes useful when it is typed. Typed output can be routed. Routed output can be validated. Validated output can be published.

Narration explains the civic state; it does not create the civic state.

Pixels are output. State is the product.

The best agent systems remember, validate, route, and trace before they improvise.

AI slop is what happens when generation crosses the commit boundary without earning trust.

Plausible is not the same as true.

Untrusted context must not become trusted instruction.

The agent earns autonomy by passing controls, not by sounding confident.

Do not start with a blank chat. Start with a repo that knows how to be helped.

Velocity comes from repeating a small disciplined loop, not from trusting a large unreviewed leap.

Write the handoff as if the next agent has never heard of this project.

Start with hazards removed. End with attention deployed. Never reverse that order.

Prompts matter when they produce artifacts the system can verify.

Review the system, not the prose.

Done means owned, tested, understood, and recoverable. Not just passing.

The session can be messy. The factory must be clean.

Stagehand gives the model a director's baton, not root access.