0xguenther/agent-write-path-runs

500 AI agent runs: duplicate writes and false success after tool timeouts. 5 models x 10 failure cases x 10 reps. Open data, harness and scoring (CC BY 4.0).

2 best points · 0 comments · 2 Hacker News posts · 1 GitHub stars · View on GitHub

Hacker News discussions

• We ran 5 LLMs 100 times each against tool calls that time out after the write (2 points, 0 comments, by 0xguenther, 2026-10-07)
• Show HN: We tested how 5 LLMs handle failed write actions (duplicate prevention) (1 points, 0 comments, by 0xguenther, 2026-10-13)