11d • Research
behind the Astra math headlines
Ten open math problems fell over the weekend to an unreleased OpenAI model. Worth reading for the method, not the headline.
The published price was under $2,000 for all ten. That figure covers only the attempts that worked — the researcher who announced it confirmed the failures twenty-nine minutes later, same thread, and nobody has said how many. Terence Tao called this exact failure mode last July: he wants AI evaluated the way aviation reports itself, cost-per-seat-mile and accident rate. OpenAI published the first and withheld the second.
The useful habit here isn't skepticism, it's structural. When a number lands, ask what's in the denominator. Then ask which half of the announcement is actually checkable — in this case the proofs genuinely are, and the process genuinely isn't.
Full breakdown, ten minutes.
0
0 comments
behind the Astra math headlines
skool.com/burstiness-and-perplexity
AI-native SEO, autonomous agents, and automation pipelines. Built for practitioners who build— not collect. Home of the Hidden State Drift Mastermind.
Powered by