From Idea to Prototype with Claude Code

Use the scoping canvas to pick one user, one job, one flow, and one AI feature

Write a kill list to keep your prototype small enough to finish in five sessions

Choose from a starter project menu if you arrive without an idea

Draft a build spec with acceptance checks, then critique it with Claude

Give Claude Code your product context and constraints in a CLAUDE.md file

Break the build into small steps with an increment plan and plan mode

Build alongside me live, committing after each working step so you can roll back safely

Prompt Claude Code for outcomes and keep each change small and testable

See a worked example from Fantartistic, a milestone art product built with AI

Connect an AI feature to your flow using your own Anthropic API key

Write down what a good output looks like before you test it

Break your prototype on purpose in the live debug clinic, then fix it with Claude

Write 10 test scenarios, run them, and log what fails

Review Claude’s output at a PM level, including security and privacy basics

Deploy your prototype to a shareable link

Run a 15-minute feedback session with 1 to 2 people using a ready-made script

Synthesize themes and rank fixes by impact and effort

Demo your prototype and leave with a prioritized roadmap