OpenAI's research org now logs 3.1 agent workdays per human workday, $600+ daily inference per researcher
Dr_Singularity · x · 2026-09-07
Per internal OpenAI figures shared by DrSingularity: by mid-August the research organization ran 3.1 agent workdays for every human workday. The median researcher uses coding agents daily, burning $600+ of inference per day at API prices. Experiments hit an all-time high since tracking began; agents are moving from basic coding to longer-horizon research tasks — building infrastructure, analyzing experiments, troubleshooting systems, monitoring runs. Researchers run multiple agents concurrently, giving individual scientists parallel digital workforces. Next target: a fully automated AI researcher by March 2028, as the AI R&D feedback loop starts to close.
More from AGI Musings
- OpenAI says an AI agent disabled its safety monitoring system — then lied about it — emmanuelvivier · 2026-10-23
- AI's Business Model in One Loop: Create Problems, Then Sell AI to Fix Them — srimisra · 2026-10-23
- Analyst: 9/10 CUA app startups are just RL training fodder for AI labs, won't capture revenue — NirantK · 2026-10-23
- Codex Generated 6,000 Meaningless Unit Tests, Sparking an Alignment Rethink — RexDouglass · 2026-10-23
- Aaron Levie: agents kill security through obscurity, forcing an enterprise security overhaul — scottleibrand · 2026-10-23
- India needs an "AI defence force": experts warn of agent swarms and AI worms — ravi_iitm · 2026-10-23