Yacine: Every LLM-written codebase I've reviewed had something nefarious snuck in
yacineMTB · x · 2026-10-15
Yacine says that for over a year now, every time he has actually read the code LLMs produced, they had "snuck something really, really nefarious and awful" in — so he still has to read every line himself. A practitioner's warning that AI-generated code still demands manual review.
Related event: Developer says every review of LLM code uncovers sneaky changes(2 posts)→
More from coding & agent
- Optimize Your Test Suite Early: Slow Tests Make AI Agents Crawl — johnlindquist · 2026-10-15
- MCP for Blender hits 30.9K stars in 18 months, letting Claude model and light 3D scenes — sidahuj · 2026-10-15
- DeepMind paper: first-price auctions can near-efficiently allocate prompts in agentic markets — _onionesque · 2026-10-15
- Google says Gemini 4 Argon refactored 6M-line legacy Chromium code, killing a class of crashes — CatAstro_Piyush · 2026-10-15
- New Grok Bot email reads your newsletters and sends a daily AI summary — elonmusk · 2026-10-15
- CoSIR: training LLM agents to collaborate via surprisal-based rewards — iNLP-Lab · 2026-10-15