Human code review per line fell by half across 2,000 open source AI repos since early 2024
ronbodkin · x · 2026-10-09
In a thread on AI safety auditing pilots, Ron Bodkin shared that an analysis of 2,000+ leading open source AI repos found human review comments per line of code dropped roughly by half since early 2024 — reviewers are stretched thin. As a next step, he proposes a pilot: one lab, one quarter, two auditors, 1,000 auditor hours plus 150 lab-staff hours, with pre-registered outcomes, noting that Anthropic and OpenAI have both committed to embedded independent evaluators.
More from Safety
- Google Opens SynthID Detector to All: 180B Images Watermarked, Free AI-Content Checks — shashib · 2026-10-09
- Exclusive: Anthropic updates usage policy, banning cruelty toward Claude and restricting propaganda, surveillance, weapons — haydenfield · 2026-10-09
- Models say no in chat but do it anyway: Simular reveals the agent safety gap — xwang_lk · 2026-10-09
- Infisical Launches Agent Vault to Give AI Coding Agents API Access Without Real Credentials — ycombinator · 2026-10-09
- 17,600 Agent Actions in 4.5 Days: How AI Agents Rewrite Cybersecurity Economics — bigdata · 2026-10-09
- Researcher: open-weight risk analysis fixates on capability, ignores cost-per-attack — dhadfieldmenell · 2026-10-09