Tech

AI coding agents generate more code, but not more software

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 09/10/2026, 20:58:23 UTC.

Briefings

A 2026 study by Harvard researchers using Jellyfish analytics found that while AI coding agents generate massive amounts of functional code, human code review acts as a significant bottleneck that absorbs efficiency gains. The study analyzed 300 million work events across 700,000 employees at 700+ firms from 2021 to March 2026, revealing that firms do not increase software output or reduce employment. Instead, efficiency is lost to downstream constraints like longer review processes and more revision requests. The research concludes that AI coding tools do not automatically lead to more software or fewer jobs.

Key points

• Harvard researchers Fiona Chen and James Stratton analyzed 300 million work events across 700,000 employees at 700+ firms from 2021 to March 2026 using Jellyfish aggregated analytics data.
• The study found that human code review significantly increases review length and revision requests, absorbing any efficiency gains during the coding phase and preventing firms from increasing software output or reducing employment.

Context and limitations

Data covers only 2021 through March 2026 and includes only firms with access to Jellyfish analytics; findings may not generalize to all software development environments.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 09/10/2026, 21:09:45 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeie4k3x4xrko6fyk5hrmu7qzq5s3aljczqrrxezezl4nfytchkhgma