OpenAI says 10,000 AI agents solved a $1 million math problem. Now mathematicians are fighting
An internal model more powerful than GPT-6 Astra produced a proposed solution to one of mathematics’ seven Millennium Prize Problems, though questions are emerging over how independently it got there.
Asanat Analysis — Why it matters
OpenAI's claim that AI agents solved a Millennium Prize Problem signals a threshold moment for machine reasoning, but the mathematician pushback reveals deeper questions about what 'solving' means in formal mathematics. If the solution relied heavily on training data, pattern matching, or required human verification of logical steps, the achievement becomes less about autonomous mathematical discovery and more about sophisticated synthesis—a distinction with major implications for how we evaluate AI capabilities in domains requiring genuine novelty.
For crypto specifically, this touches the ongoing debate about on-chain AI agents and autonomous smart contracts. The incident demonstrates that scaling compute and agent count doesn't automatically yield trustworthy, independent problem-solving. Similar skepticism will dog proposals for fully autonomous DeFi protocols or AI-driven risk management systems. Mathematicians' scrutiny here foreshadows the due diligence bar for deploying AI agents in financial systems where opacity around reasoning paths creates unacceptable counterparty risk.
The framing of 10,000 agents working in parallel also echoes distributed ledger logic—but highlights why crypto systems still require formal verification and human governance layers. The gap between 'produces an answer' and 'proves the answer reliably' is where trust is built or lost.