Related video: Anthropic mathematician claims Claude helped prove Riemann Hypothesis in six weeks, @TheAIGRID
OpenAI releases model perplexity scores, proposes pacing limits on AI advances
Techmeme · · 20-second read
Top comments from YouTube.
November 22, 2026 · Emerging AI
OpenAI has published perplexity scores—a measure of how well its language models predict data—and proposed using such metrics to pace AI development. The company flagged challenges in comparing performance across different research labs and noted that standard metrics don't capture all kinds of advances, particularly those outside of pretraining.
OpenAI warned of a possible "avalanche" of AI improvements around 2027, suggesting that pacing frameworks based on measurable performance thresholds could help manage the speed of capability gains.
This is an AI-written summary. The original reporting belongs to Techmeme. How we write these · Spot a mistake?
Get the AI brief every morning.
The five AI stories that matter each morning: startups, money, jobs, tech and what's coming next. Free, and one click to unsubscribe. See our Privacy Policy.