Deferred items

This workshop is a tool comparison, so it stops before turning Snake into a production app. The next useful step is to repeat the same task with your own constraints and compare the results.

Repeat the same task across tools

Use one fixed prompt and one fixed acceptance check. The workshop prompt is:

implement snake in react

Then compare each result with the same local commands:

npm install
npm run dev
npm run lint
npm run build

That keeps the comparison honest. If one tool creates a great preview but the exported project cannot build, that is part of the result. If another tool creates plain code that builds cleanly, that is also part of the result.

Add tests

The live demo does not add tests. After the workshop you can ask an IDE assistant to add tests around the pure game rules:

• moving one cell in the current direction
• blocking direct reversal
• eating food and growing the snake
• detecting wall collisions
• wrapping in pass-through mode

For this app, the first refactor would be to move game-state transitions out of the React component. That gives the tests plain functions to call instead of making every test drive keyboard events and timers.

Compare generated structure

The two local apps already show a structural difference. snake-chatgpt keeps the game in src/App.jsx. snake-claude-code keeps the page shell in src/App.jsx and moves the game into src/components/SnakeGame.jsx.

Use that as a review checklist:

• Does the tool separate page shell from game logic?
• Does it leave unused Vite starter CSS behind?
• Does it use dependencies that are already installed?
• Does it add features in one place or spread changes across many files?
• Does it leave you with commands that pass?

End-to-end app follow-up

After the workshop you can turn the comparison into a larger app:

• Use Lovable to bootstrap a React frontend.
• Export the project.
• Continue in Cursor or VS Code.
• Add a Python backend.

For the Snake game, a backend could support a leaderboard or saved games.

That build is bigger than this workshop because it needs an API contract, persistence, deployment, and tests. Keep it separate instead of adding it to the comparison demo.

Agent follow-up

The workshop also points to building your own agents. Two concrete follow-up ideas are an agent for Django-based projects and an agent that uses the GitHub API.

Use the RAG agents workshop and LLM Zoomcamp as the next materials if you want to implement tool calling directly.