Does AI Actually Boost Developer Productivity? (100k Devs Study)
summary
Stanford analyzes nearly 100,000 developers across hundreds of companies and finds AI yields an average productivity boost of ~20% but with wide variance and some teams seeing decreases. Method measures functionality per commit from private git history using an automated model, surfaces ghost engineers ~10%, and reports a 15–40% uplift on low‑complexity greenfield work versus 0–10% on high‑complexity brownfield projects. Gains fall as codebase size and context length grow and large models already drop performance past 32k tokens.
AI-written summary · about
chapters (8)
- 0:00Introduction and the context of AI in software development, including Mark Zuckerberg's bold claims.
- 4:37Limitations of existing studies on AI's impact on developer productivity.
- 7:19The methodology used by the Stanford research group to measure productivity.
- 9:50The overall impact of AI on developer productivity, including the concept of "rework."
- 11:42How productivity gains vary by task complexity and project maturity (Greenfield vs. Brownfield).
- 14:21The impact of programming language popularity on AI's effectiveness.
- 15:42How codebase size affects AI-driven productivity gains.
- 17:22The final conclusions of the study.
related talks
18mHumans + AI: The Future of Software Development Teams
46mMeasuring the Impact of AI on Developer Productivity
12mHow is AI reshaping tech teams
27mThe state of AI in software development: Insights across 400+ organizations
44mSpeed Is Easy, Trust Is Hard: AI's Real Impact on Developer Productivity
23mThe AI productivity paradox: Why feeling faster isn't the same as going faster
22mThe AI Productivity Paradox in Engineering
20mThe Specialisation Trap