OpenAI's first research-acceleration report: AI spend per researcher up ~100x, calls for mandatory RSI reporting
kliu128 · x · 2026-11-19
OpenAI released a new transparency report on how AI is changing its research. Key findings: researchers' AI use is rising exponentially and agents complete longer tasks with steadily improving zero-intervention success rates, while human involvement remains central. Median daily spend per researcher on Codex/ChatGPT Work grew from near zero in January 2026 to $2.4K by October; the 90th percentile exceeds $20K. In a survey of 101 researchers, the median would trade 40% of team headcount or 30% of team compute to restore access to coding agents better than GPT-5.2 for three months (50% of compute if asked in Jan 2027). Time spent on safety and alignment is growing faster than any other research area. OpenAI argues AI labs should be required to publicly report indicators of recursive self-improvement (RSI).
More from AGI Musings
- Anthropic insider: if superintelligence arrives, the company expects de facto nationalization — teortaxesTex · 2026-11-19
- Commentary: Claude's inner alignment is fine — Anthropic's guardrails are the problem — repligate · 2026-11-19
- 'Open source AI believers are like libertarians': a spicy take on grifters — zetalyrae · 2026-11-19
- Gary Marcus slams fully handing AI research to AI as the dumbest idea in human history — GaryMarcus · 2026-11-19
- Christianos interview sparks alarm: 'state of emergency' AI scenario decried as coup justification — joshua_saxe · 2026-11-19
- Prediction: 2027 could see 1000x more AI-driven scientific breakthroughs — Dr_Singularity · 2026-11-19