Problem
Need a safe space for AI to experiment, test assumptions, and explore libraries without affecting production code or committing throwaway experiments.
Pattern
Allow it to experiment when it gets stuck.
For example, create an isolated playground folder (can make it .gitignored) where AI can:
- Test library behaviors and discover constraints
- Validate assumptions about APIs
- Try different approaches without consequences
- Build proof-of-concepts before real implementation and gain valuable learning
When AI gets stuck or when working with new libraries or uncommon languages:
- Pull back from complex debugging
- Have AI experiment in playground
- Test assumptions directly
- Discover library constraints quickly
Example
Building chess app, nothing working. Instead of debugging through UI → components → chess.js integration, stopped and had AI write playground script to test chess.js directly.
Quickly discovered: chess.js requires kings on board. Would have taken much longer debugging from the top down.
