to-prototype
Prototype
Answer one decision-changing question with throwaway code. Prototype code never graduates: the production task rebuilds the chosen behavior under its own acceptance contract. A prototype yields evidence and a decision, not a feature.
1. State the question
Name the unknown, the observation that would answer it, and the time or scope limit. Read the code and docs first; a question they already answer needs no prototype. Pick the smallest form:
- A visual decision: references/ui.md.
- Logic or state behavior a person must explore: a standalone HTML demo, references/logic.md.
- A technical property such as integration behavior, latency, or library support: a small script or harness that records input, output, and measurement, with no UI.
2. Run it
Work in a scratch location or local worktree, name it as a prototype, and give one command or path to run it. Use memory-backed or scratch data; touch real services only when authorized and necessary. Skip production scaffolding and polish, but keep enough checks and error output to trust the result. Record observations and failed attempts. Stop at the agreed limit even if the question stays open.
3. Return the decision
Return: