Voices
Orosz Argues Code Review Will Break at Most Startups: Agents Produce Far More PRs, Far More Verbose, and Nobody Wants to Review AI Code
Citing a new Linear report, Orosz posted on Aug 17 that 'coding agents (obviously) mean a LOT more PRs. And more verbose PRs. It's already less motivation to review AI-generated code… and so much of it?!!' The claim tracks with LinearB's separately published 2026 benchmarks over 8.1M PRs, which found agentic AI PRs sit in review 5.3× longer than unassisted ones and that only 32.7% of AI-assisted PRs merge within 30 days versus 84.5% of manual ones. For builders the implication is that review capacity, not generation, is the binding constraint in agentic workflows.
Source
↳ Follow the thread