Under the leadership of Tianyi Peng, a distinguished alumnus of Tsinghua University's renowned Yao Class, Anthropic's artificial intelligence model, Claude, has achieved a groundbreaking milestone by completing the first end-to-end, fully computer-verifiable formal proof of Fermat's Last Theorem within a mere 11 days. This monumental proof, which spans approximately 13 million lines of Lean code and encompasses over 30,000 intermediate theorems, boasts a code size that exceeds fivefold that of Lean's core mathematical library, Mathlib. The endeavor required the meticulous transformation of a human-readable proof into a formal proof that could be rigorously verified line-by-line by a computer.

Throughout this intricate process, multi-agent collaboration initially encountered some confusion. However, this was swiftly streamlined through the utilization of the Prove2Me platform, with human intervention limited to providing only minimal high-level guidance. In other news, OpenAI is progressively rolling out its latest iteration, GPT-6 Astra, to paid ChatGPT users, offering significant enhancements in capabilities, particularly in handling long-chain agent tasks.