January 5, 2026
Lint cops vs. AI bots
How Twitch tamed a million lines of TypeScript
Twitch’s million-line “fix” meets AI slop outrage
TLDR: Twitch used a simple snapshot trick to track rule exceptions and auto-alert maintainers in a huge codebase. Commenters mostly blasted the post as AI-written “slop” and outdated, highlighting a growing backlash against robotic corporate tech blogs while the idea itself quietly earns respect.
Twitch says it corralled a million lines of code by turning rule‑breaking into a list that the core team owns — no hard blocks, just visibility and conversation. In plain terms: instead of hiding “ignore this rule” comments inside files, they run a check that pretends those don’t exist, save the results to a snapshot, and auto‑pull the standards team into reviews when exceptions move or multiply. It’s a soft‑touch system for a giant one‑codebase world (a “monolith”) with hundreds of developers, explained in a post that nods to the older Twitch engineering blog.
The comments, though? Absolute meltdown. Readers piled on with “AI slop” callouts, accusing the write‑up of being machine‑generated and boring. One joked they should just post the prompt instead of the output; another sneered it’s old, mundane history. The meme of the day: “lint cops vs. AI bots,” with folks dunking on corporate blog tone as suspiciously robotic. A few calmer voices noted this is a practical, non‑blocking way to keep standards visible, but the hot take tide was clear: the method is neat, the story feels AI‑flavored — and the audience is over it. Reminder: LLM means “large language model,” which fuels the AI jokes here. Drama level: high.
Key Points
- •Twitch managed standards for a million-line TypeScript monolith used by around 300 developers.
- •Unintentional bypassing of standards accumulated via eslint-disable comments, copied exceptions, refactors, and temporary escapes.
- •The team aimed for incremental evolution, visibility of exceptions, and a feedback loop without blocking development.
- •They implemented ESLint snapshots that ignore inline disables, stripped line numbers, and used CODEOWNERS to enforce reviews.
- •The system made exceptions visible, enabling explicit discussion and adaptation of rules; a Twitch blog post offers further technical detail.