Adopt a structured, software engineering-inspired approach to iterating on AI designs. Use forking to safely explore new ideas and apply a two-step debugging prompt to reliably fix errors without frustrating back-and-forth.
with Claire Vo
When you have a working version of a prototype (a 'baseline'), create a duplicate or 'fork' before trying to add a new feature or make a change. This allows you to explore ideas in parallel without breaking your stable starting point.
Keep your project organized by clearly naming your forks. For example: Baseline - Profile Page, Var 1 - Add Edit Button, Var 2 - New Layout. This makes it easy to track different design directions.
When the AI doesn't follow an instruction correctly (e.g., fails to delete an element), first ask it to diagnose the problem *without* writing any code. This forces the AI to analyze the root cause.
Tried to delete this but it didn't work. Can you explain why? Don't write any code.After the AI has explained the 'why' behind the problem, you can then ask it to implement the fix. This 'explain, then solve' pattern leads to much more reliable and accurate results when debugging.
Use an AI agent to automatically catalog physical items like toys and books from photos, creating a searchable inventory in a note-taking app. The agent can then connect these physical items to digital plans, like suggesting relevant toys for a lesson.
Partner with an AI coding agent to build and deploy a custom TV application from scratch, even with zero prior coding knowledge. This workflow allows you to create a safe, curated content environment for your family by pulling from approved sources.
Use an AI agent to digitize curriculum books from photos, automatically generate structured lesson plans, and create custom learning materials like watercolor illustrations. This saves hours of manual data entry and creative work.
Join 100,000+ product managers who use ChatPRD to write better docs, align teams faster, and build products users love.