Agents All the Way Down: How AI Coding Agents Changed How I Write Django
summary
Ten Django apps, a 19‑year‑old legacy codebase, and a year of using Claude Code/Codex agents lead to concrete changes: a bug triaged and patched from a phone during a Disney trip, a primary tract foreign key added with a data migration to replace a brittle boolean pattern, and deliberate work to get a reduced test runtime so agents can verify fixes. Agents excel at framework conventions but fail on migrations, multi‑app relationships, and over‑mocking, so scaffolding is kept lightweight and disposable rather than permanent.
AI-written summary · about
chapters (11)
- 0:00Introduction and vacation fix
- 4:33Background and company context
- 7:05Evolution of AI coding tools
- 9:46Django conventions and agent alignment
- 13:26Refactoring primary tract model
- 15:18Understanding slop in agent code
- 18:21Reviewing agent changes
- 20:08Personal projects and trade-offs
- 22:34Cognitive debt and learning
- 23:48Judgment and data structures
- 24:11Closing and Fred Brooks quote
