Unverified50% confidenceEventExact time
A tech enthusiast who had previously participated in DEF CON puzzle challenges conducted a real-world test using a new model called "Astra"
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/12/2026
First Seen
Valid until: 12/11/2026
Sources
Related Entities
Related Claims
UnverifiedSome users use Fable as a validator for Astra's solutions, having one model plan and the other review.63% similarUnverifiedWhen solving the puzzle, Astra used the official hints provided by the puzzle creator, which were the same hints given to human contestants at the time63% similarUnverifiedAstra's main agent proposes hypotheses or theories, then dispatches sub-agents to actually verify them and observe the running results62% similarUnverifiedAstra模型探索了开放的研究问题，生成了新的证明或更强的数学界限，并协助将论证整理为可供审阅的论文手稿60% similarUnverifiedNext-generation evaluations such as FrontierMath and ARC Prize attempt to use novel, unpublished difficult problems to test models' genuine reasoning capabilities.59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/914322API
curl https://kongchang.com/api/v1/knowledge/claims/914322MCP
get_claim(id=914322)