This site is a work in progress and has not been widely shared. Content may contain errors. Feedback is welcome.
This site is undergoing review. Some annotations were human-generated, some AI-generated — all are being verified.
Back to datasets

METR Open-Source Developers

Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity

AI-focusedPublicWorker-side
Visit Dataset
Specific Type
Experimental AI usage evidence
Dataset Type
Cross-sectional
Institution
METR
Institution Type
Non-profit
Level of Focus
Individual
Most Granular Level
Individual developer level
Perspective
Worker-side
Time Coverage
2025
Frequency
One-time experiment
Sample Size
16 experienced open-source developers (small sample)
Geographic Detail
Not specified
Occupational Classification
Software developers
Industrial Classification
Technology/Software
Other Classification
Experience level; repository size
Key Variables
Task completion time; productivity measures; developer perceptions; objective vs subjective performance
AI/Tech Tracking
Cursor Pro with Claude 3.5/3.7 Sonnet usage
Access Details
Available from METR
Notes
AI tools slowed developers by 19%; developers incorrectly believed AI sped them up by 20%; 246 real GitHub issues across large repositories

Key Papers

METR (2025)