← Learn
02 · Method & Knowledge 4 lessons · 1 min read Learned

Coding Workflow With AI

Structure context, invariants, incremental steps, and stop conditions so every AI-assisted code change stays bounded and verifiable.

Best for: Vibe coders and developers reducing AI rework. Deliverable: A bounded task spec with real verification.

#AI Coding#Vibe Coding#Verification

Ground the Codebase Before Requesting Edits

Before asking for changes, state what the app does, how it behaves today, which files own the logic, and what must remain untouched.

Key points

  • Current behavior
  • Target behavior
  • Non-negotiable invariants

Exercise Pick a real code change and describe current state, target state, and boundaries in three sentences.

Slice Tasks Into Observable Steps

Break “add search” into query input, matching results, empty-state reset, and URL persistence. Ship the main path first without opportunistic rewrites.

Key points

  • Concrete user trigger
  • Observable state change
  • Explicit failure/empty state

Exercise Rewrite a vague feature wish into three executable acceptance steps.

Verify From the Real Entry Point

Generated files are not proof of working UX. Walk the real user entry point, inspect runtime behavior, and separate observed errors from guesses.

Key points

  • Test via actual UI/API entry
  • Record exact reproduction steps
  • Re-check adjacent flows

Exercise Execute your three acceptance checks, log real outcomes, and mark anything untested honestly.

Define Clean Stop Conditions

Stop once target behavior passes verification and out-of-scope ideas are logged for future iterations. Fold reusable patterns back into your notes.

Key points

  • Target achieved
  • Affected paths verified
  • Deferred items documented

Exercise Write a handoff note stating exact verification—banning vague words like “optimized.”

Finished this topic?

Topic learned 🎉

Once you have read the lessons and tried the exercises, mark it — you move one step forward on the race.

Race position is now 0%

Next topic →

Sources:AI Coding Notes · Output-Driven Learning · Learning Route Design