Fixit linter+AI coding
Talkat PythonAsia 2026 · on Saturday, 2026-03-21· at 14:45· Teresa Yuchengco Auditorium (Main Hall)·
summary
Claims that AI can generate working lint rules are demonstrated: an AI prompt produces Fixit + LibCST rules that migrate imports and calls from logging to structlog, Fixit runs find-and-fix changes while preserving comments and whitespace, and its built-in test harness checks valid/invalid cases. A mutation tool mutoto mutation testing runs against those tests, reporting e.g. 12 killed and 3 survived for an 80% mutation score, and the whole flow runs in GitHub Actions as an automatic CI guardrail.
AI-written summary · about
chapters (10)
- 0:00Fixit linter and AI coding
- 1:32Custom linters in AI era
- 6:56Fixit framework for linters
- 9:50CST and Fixit role
- 14:17AI-generated lint rules
- 16:26Logging to structlog migration
- 19:24Guarding AI code with linters
- 21:58Mutation testing lint rules
- 26:01CI/CD tool integration
- 27:05Summary
related talks
07mMake Linting Great Again
19mMake Linting Great Again
36mCode Mending in the AI age
27mRules_lint: Formatting and Linting All Languages
34mLint to the Rescue
43mHow to Use AI to Build Accessible Products
22mIncreasing code habitability by using AI coding assistants
29mOne-Click Code Fix: Securing Code Using AI