Google expands multi-agent research workflows

What happened: Google Research described Antigravity’s /teamwork approach, with specialised agents, restricted task context, and critics or auditors checking results from fresh contexts. The company reports applications across mathematics, software engineering and science in collaboration with human experts. It also says applications are open for 10,000 academic researchers to receive six months of its highest usage tier.

Why it matters: Separating production from verification is more useful than simply increasing the number of agents. For research and engineering teams, the design suggests a practical way to reduce shared blind spots and preserve focus during long tasks.

Caveat: These are Google-reported results. Individual scientific and performance claims require scrutiny of the underlying evidence. Some specialised workflows remain forthcoming; not every capability described is generally available.

Dates: Report and access-program announcement published 13 November 2026. The reported collaborations took place over preceding weeks, not solely on publication day.