Stanford researchers analyzed 100,000+ developers across 600+ companies to measure AI's impact on coding productivity. The study found AI increases productivity by 15-20% on average, but effectiveness varies significantly based on task complexity, codebase maturity, and programming language popularity. AI performs best on simple, greenfield tasks in popular languages, showing 30-40% gains, but can actually decrease productivity on complex tasks in obscure languages or large codebases. The research challenges vendor-led studies that focus on commit counts, revealing that AI often generates additional rework and bug fixes that offset initial productivity gains.

18m watch time

Sort: