GPT-5.6 Sol Ultra Claims Proof of 50-Year Graph Theory Conjecture in Under an Hour
One-line summary: OpenAI announced that GPT-5.6 Sol Ultra produced a proof of the Cycle Double Cover Conjecture—a 50-year-old open problem in graph theory—by running 64 parallel subagents for under an hour.
Key facts
- The Cycle Double Cover (CDC) Conjecture was independently posed by Szekeres (1973) and Seymour (1979); it asks whether every bridgeless graph has a set of cycles in which each edge appears exactly twice
- GPT-5.6 Sol Ultra ran 64 subagents in parallel and produced a candidate proof in roughly 60 minutes
- OpenAI published the complete 700-word orchestration prompt, sharing the multi-agent coordination techniques used for ultra-complex reasoning tasks
- Caveat: the proof has not yet passed independent mathematical peer review; the CDC has attracted flawed proofs in the past that were later retracted
Why it matters
If verified by the mathematics community, this would mark the first time an AI system resolved a major open conjecture through multi-agent collaboration—fundamentally reshaping expectations about AI's role in formal scientific discovery.
Read more
- GPT-5.6 Sol Ultra reportedly solves 50-year-old math problem — The Decoder
- GPT-5.6 Sol Ultra Proves the Cycle Double Cover Conjecture — CryptoBriefing