AI Researcher: LLMs Handle Negation Poorly — 'Don't Do X' Prompts Have Deleted Code
mmitchell_ai · x · 2026-10-23
AI researcher Margaret Mitchell (@mmitchellai) explains why user instructions often fail.
- Negation trap: saying "don't do X" doesn't work as well as hoped because LLMs are weak at processing negation — she cites multiple real cases where this wiped out code.
- Why users are vague: people are generally bad at writing instructions, and without extensive experience with computational systems, they don't know how to break down their language to elicit reliable outputs.
- Practical takeaway: prefer positive descriptions of desired behavior over negative constraints in prompts.
More from coding & agent
- Claude Code Pushes Urgent Security Update Across All Installations — umeshonai · 2026-10-23
- Claude cut a 2.6-hour video export to 20 minutes — but only after being told to rebuild the server — byhewar · 2026-10-23
- 3 flight-booking rules my AI agent checks that I'd never do by hand — Efistoffeles · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23