Advancing science & engineering with Antigravity’s /teamwork

判断

Research leads at Google and external partner labs must decide whether to route open-problem work through Antigravity /teamwork now, accepting workflow rebuild and strict context partitioning in exchange for parallel exploration and independent falsification. The call is theirs because Mirrokni says tools are available externally while more features are still coming.

依据

Vahab Mirrokni, Google Fellow and VP, Google Research, says teams across Google and academic partners used Gemini 4 Argon, Carbon, Hydrogen and Antigravity’s /teamwork multi-agent workflows on open problems in mathematics, chemistry, systems software and ML engineering. He says /teamwork addresses two limits: unpredictable generation, by pairing parallel exploration with dedicated falsifiers such as the Lean 4 proof checker, unmodified software test suites, execution simulators, web search or hardware counters; and attention loss over multi-day tasks, by partitioning context so each specialized agent receives only the files and tools for its step, with coordinators handing off to fresh successors and critics/auditors verifying from clean contexts. The source cites the Courtade–Kumar conjecture, open since 2013, proved in full generality and formalized in Lean 4 with 50 million lines of code, plus a 14-variable optimization reduction. Mirrokni says the tools are available externally, with more features coming in the coming weeks.