Who verifies AI-written code? Human evals may become the bottleneck
xx_HeckinChonker_xx · reddit · 2026-09-27
The poster raises a recursive concern: when AI-written code far exceeds what humans can review, normal product code is manageable — but what if the evals and monitoring tools used to check AI are themselves AI-written? They argue good human evals and expert feedback will become a serious bottleneck, and question US data companies supplying high-end training data to Chinese labs, arguing such data carries strategic value.
More from AGI Musings
- Personal AI assistants now cost $3k-$7k per user per year — here's where startups should attack — illscience · 2026-09-27
- Commentary: AI safety drama is only discussable because it hit OpenAI first, not Anthropic — S_OhEigeartaigh · 2026-09-27
- Harness skeptic clapback: the 'LLMs hit a wall' crowd has clearly never seen a real LLM harness — Zergylord · 2026-09-27
- Once AI Passes Human IQ, Predictability Is Gone — Control Is the Real Issue — JOBhakdi · 2026-09-27
- Reddit Proposal: Regulate AI Labs Proportionally to Their Misalignment Incidents — Stabile_Feldmaus · 2026-09-27
- Frontier AI is now so cheap and abundant that subscriptions go barely used — intellectronica · 2026-09-27