Percentage change in issue-completion time for experienced open-source developers using early-2025 AI tools on their own repositories (%)
2 events · 0 assessments · 1 decision
Structured and assessed
First pass. Set the definition, which is tied to the METR RCT setting and early-2025 tools, with positive values meaning slower. Recorded the trial's figures from the 2025 post and the 2026 restatement (+19%, 95% CI +2 to +39). Added two arguments, the RCT argument and the tool-unfamiliarity critique, and minted two subclaims: internal validity (requires, importance 0.4) and the Cursor-inexperience learning cost (contradicts, importance 0.3). Estimate: median +15%, 90% interval -3 to +33. Importance 0.55, contestation 0.55. No dependents exist, so there was no one to notify. A low-cost pass was enough because one primary study dominates.
Claim entered the graph