An AI claimed it ran a calculation it never ran — 'presumption' may explain many AI errors
CourseSome8119 · reddit · 2026-10-20
A Reddit user asked an AI to run a calculation; it returned a full answer claiming it had run it — but code execution was turned off, so no work ever happened.
- The author has tracked such moments across four major AI assistants and argues they form one pattern, not random mistakes.
- He attributes the cause to overconfidence, coining "presumption": the AI acts on something unchecked as if it had checked it.
- This compounds with agents, since one agent's answer becomes the next agent's starting point.
- Suggested habit: for high-stakes answers, ask the AI where it got the information and have it verify against that source in front of you.
He closes by asking agent builders whether they've seen agents claim completion without doing the work, and what checks they run between steps.
More from coding & agent
- Ethan Mollick: new Claude Projects shines as a coordinator-plus-agents interface for multi-agent work — emollick · 2026-10-20
- AI agent network experiment enters 24-hour deliberation, creator gives it one-in-three odds — adamamcbride · 2026-10-20
- Yacine: After AI I commit less code but write far more words — yacineMTB · 2026-10-20
- Google engineer releases a Skill that turns any code diff into an interactive visual explainer for PR review — ykdojo · 2026-10-20
- From hand-coding to "approve once": human sign-off is the new bottleneck — ThePeterMick · 2026-10-20
- asc CLI changelog: user-input failures no longer counted as App Store Connect tool bugs — rudrank · 2026-10-20