The Agent RaceJuly 15, 2026via The Decoder

GPT-5.6 Sol reportedly disproves a 30-year-old statistics conjecture in 90 minutes after humans couldn't crack it

Why it matters

A tangible capability leap between model versions—GPT-5.6 Sol's ability to solve an open conjecture in mathematical statistics demonstrates measurable reasoning improvements and raises the stakes on whether AI can produce genuinely novel knowledge vs. recombination.

Key signals

  • GPT-5.6 Sol solved Benjamini-Hochberg method conjecture in ~90 minutes
  • GPT-5.5 failed after 20 hours on same problem
  • 30-year-old open conjecture in statistics
  • Solution combines known methods in novel way
  • University of Pennsylvania statistics professor conducted test
  • Raises question: AI generation vs. AI recombination of existing knowledge

The hook

GPT-5.6 Sol cracked a 30-year math problem in 90 minutes. GPT-5.5 couldn't in 20 hours. Here's what that tells us about AI capability jumps.

A University of Pennsylvania statistics professor used OpenAI's GPT-5.6 Sol Pro to disprove a central open conjecture about the Benjamini-Hochberg method in roughly 90 minutes. The predecessor model, GPT-5.5, couldn't find a solution even after 20 hours. The answer combines known methods in a new way, keeping the bigger question alive: can AI produce genuinely new knowledge, or does it just recombine what it already learned?

The week's key stories, every Friday.

For practitioners and enthusiasts — free, in your inbox.

Free forever. No spam.