What do we do now?
summary
Built Solo entirely with AI using Tari (Rust) + TypeScript, the project reaches 3500 commits and shows about 25% fixes of those commits as followups. The architecture centralizes business logic into a canonical ops layer (MCP + webview adapters), adds deterministic pest and eslint checks, a custom end‑to‑end harness, and replaces xterm with a bespoke Rust terminal. The result is production releases guarded by ~14,000 tests and automated CI gates that largely work.
AI-written summary · about
chapters (16)
- 0:00Introduction and background
- 1:58AI and developer value
- 4:45Moving up in engineering
- 5:12First draft with AI
- 8:30Fixing mistakes and engineering judgment
- 11:30Building boxes for AI
- 13:08Centralizing operations layer
- 15:52Handling AI mistakes deterministically
- 18:36Locking in rules with tests
- 20:52Baking institutional knowledge
- 23:21Testing everything thoroughly
- 25:23Rebuilding and engineering decisions
- 28:48Leverage through prototypes
- 31:54Discovering system shape
- 33:30Judgment and system design
- 35:51Conclusion and future
related talks
28mRust Irgendwie, Irgendwo, Irgendwann
21mHow to Start a Solo Project that You'll Stick With
33mProject Update: Lang Team
40mRust at 15%: production-ready without the pain
21mModernizing Our Oldest, Scariest Code with AI
20mMastering Rust on Edge and Cloud — AI/IoT Case Studies
14mDeveloper productivity in a post-AI world
38mRust on the frontend