Codemods in the Era of AI
summary
Describes converting hundreds of React class components with a repeatable algorithm, using JSCodeShift codemods to run across thousands of files, handoff to agents/CLI tools for PRs and CI fixes, and manual work for the risky long tail; reports converting roughly 70% of components (about 550 of ~700), shrinking ~10,000 class methods down to ~2,000 and tackling a separate modularization of a client with ~10,000 files by splitting big utility barrels into single-export modules.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:31Web migration nightmares
- 4:23AI coding tools evolution
- 10:48Code and AI mental models
- 15:30Converting React class components
- 22:14Automating migrations with codemods
- 26:46Modularizing tangled codebase
- 32:26Future of AI-assisted migrations
related talks
40mRefactoring large JavaScript codebases
27mHow Codemods Turned a Nightmare into a 3-Minute Migration
25mManage large refactors with JSCodeShift
28mCross-File Codemodding
21mAutomating Front-End Refactoring
22mMaking Code Automations at Scale: Leveraging AI for Code Migrations
47mComputer, refactor my code!
34mCodemods and Transforming JavaScript Code