Back to all posts

Launching Habagou

Shipping a Chinese handwriting practice app with an agent-driven development loop.

projectsai-assistedlearning

Today I’m launching Habagou, named for the Chinese word for pug: 哈巴狗 (hābāgǒu). It’s a web app for practicing how to write Chinese characters, one stroke at a time. You can try it in your browser or read the source on GitHub.

The interesting part was the shipping process:

  • Claude Design generated screenshots and mocks to make the product feel concrete before implementation.
  • Fable turned features—such as adding authentication—into plans and GitHub tickets.
  • Fable and GPT 5.5 drove ticket execution with parallel subagents where appropriate. I used the built-in browsers in both Claude and Codex Desktop to keep the loop moving; some runs worked independently for more than an hour.
  • Fly.io, Neon, and Auth0 got the app into production without turning the homelab into the hosting platform.

The through-line was simple: use the agents to make the planning, delegation, and verification loops shorter, then keep the goal fixed on shipping a real product.


This post was drafted in collaboration with GPT 5.6-terra. Attributing when AI is part of the brainstorming or writing process and tagging those posts w/ ai-assisted.