MIT study. Code volume surges by 300%, but output increases by only 30%: The AI dividend meets an awkward reality.
Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-26
An MIT study of over 100,000 GitHub developers finds that AI coding agents raise commit volume by up to 180% but increase actual software releases by only 30%, with an estimated labor-substitution elasticity of just 0.25 due to persistent human bottlenecks in review, testing, and shipping.
Appears in
Extraction
Topics: ai-coding-toolssoftware-productivityai-economydeveloper-tools
Claims
- AI coding agents increased GitHub commits by 180% but software releases rose only 30%, revealing a large gap between code production and shipped software.
- Autocomplete raised commits by 40%, interactive coding agents by 140%, and autonomous coding agents by 180%, while releases grew at most 30% across all generations.
- App marketplaces saw more new apps but no increase in total usage, suggesting AI-generated software is not being adopted by users.
- The estimated elasticity of substitution between AI and human coding labor is 0.25, meaning large AI capability improvements displace only a small fraction of human work.
- Software production has 'weak links' — faster code generation provides limited end-to-end benefit when humans must still review, test, connect, and ship the work.
Key quotes
Autonomous AI coding agents raised commits by 180%, but releases rose only 30%.
The paper's main idea is that software production has weak links, so faster code writing does not help as much when humans still need to review, connect, test, package, and ship the work.
The estimated 'elasticity of substitution' is 0.25 i.e. for every big improvement in AI's usefulness, only a small amount of human work can be replaced.