In Google Antigravity, we not too long ago launched plenty of updates to Teamwork, a framework that permits autonomous groups of AI brokers to collaborate, critique, and iterate over hours or days to unravel complicated, long-horizon challenges. Pairing Gemini 3.7 Flash with this multi-agent orchestration accelerated downside fixing throughout analysis and engineering:
- Math and theoretical laptop science: Solved seven open issues throughout prime venues (FOCS, JMLR) — together with Knuth’s Cycles Conjecture (verified in Lean with 40+ web page proofs), sparse convex optimization, provable LLM quantization, and prefix-matrix factorizations — whereas reaching 71% on TCSBench.
- Methods engineering: Constructed a cycle-accurate, out-of-order RISC-V CPU simulator from scratch that boots the xv6 working system to shell with 0.71% cycle alignment error in opposition to {hardware} floor reality.
- Open-source software program: Landed efficiency optimizations upstream in core libraries, together with Eigen (SIMD fast-paths) and ParlayHash (2x insert throughput, 25% reminiscence discount).
Examine all of the wins on the Antigravity blog.
