AI Coding Output Up, Quality Stable
Benchmarks and industry studies show teams using AI coding tools heavily produce roughly twice as many weekly pull requests (2.2 vs 1.12 PRs per engineer) with no significant hit to code quality—a net boost to iteration speed. The findings are backed by a Business Insider report and an industry benchmark from Jellyfish in partnership with academic researchers. (businessinsider.com) (radicaldatascience.wordpress.com)
Jellyfish’s public benchmark draws from more than 700 companies, roughly 200,000 engineers and about 20 million pull requests analyzed to measure AI’s real-world effects on engineering teams. (jellyfish.co (jellyfish.co)) The study reports a median of 63% of engineers using AI tools at least weekly and finds that 64% of companies now generate a majority of their code with AI assistance. (platformengineering.org (platformengineering.org)) Jellyfish’s research shows AI-assisted pull requests are about 18% larger on average, with the size growth attributed mainly to more verbose code rather than broader scope changes. ) The benchmark also notes teams with deeper AI integration see roughly 24% faster cycle times on average, while adoption benefits vary with code architecture—centralized codebases seeing far larger gains than highly distributed ones. (platformengineering.org (platformengineering.org)) Autonomous agent activity remains a small but accelerating slice of activity, with leading companies already reporting autonomous agents accounting for up to 10% of PRs and exponential growth in agent-driven workflows. (platformengineering.org (platformengineering.org)) Jellyfish documented these findings in partnership with academic researchers and announced collaborations with Harvard Economics, OpenAI and Augment Code to validate methodology and bring vendor-level AI telemetry into its benchmarks. (jellyfish.co (jellyfish.co)) Business Insider framed the wave of results as a rapid productivity surge while reporting that multiple industry leaders and researchers described the change as a “rocket ship,” citing Jellyfish’s dataset and analysis in its coverage. (businessinsider.com (businessinsider.com))
