# Antigravity /teamwork — Google's agent teams prove an open math conjecture

> Google Research reports what Antigravity's multi-agent /teamwork system has solved, and opens free access for 10,000 academics.

- Org: Google Research
- Date: 2026-11-13
- Category: showcase, tool
- Source: https://research.google/blog/advancing-science-engineering-with-antigravitys-teamwork-feature/
- Canonical page: https://ai-tldr.dev/releases/google-antigravity-teamwork-research-results/

## In short

Google Research says Antigravity's /teamwork agents, working with human experts, proved the 2013 Courtade–Kumar conjecture in full and checked it in Lean 4. Google offers 10,000 academic researchers six months of free top-tier access.

## What it is

Google Research has published a set of results from /teamwork, the multi-agent system inside Google's Antigravity coding tool. The headline result is a full proof of the 2013 Courtade–Kumar conjecture from information theory, made with human experts and formalized in Lean 4 as 50 million lines of code.

## How it works

Several specialized agents explore a problem in parallel, and automatic checkers such as the Lean 4 proof checker or a project's own test suites decide which work is correct. Each agent gets only the files and tools it needs for its step, and the system picks a reusable pattern, such as Long Proof or iterative coding, for each prompt.

## Why it matters

The post moves agent teams from coding demos toward checked research results in math and systems software. Academics can apply for six months of free access at the highest usage tier, which lowers the cost of trying long multi-agent runs on their own open problems.

## Who it's for

mathematicians, CS researchers, developers using Antigravity

## Try it

Run /teamwork or /boost in Google Antigravity

## Quick facts

- Maker: Google Research, Antigravity team
- Feature: /teamwork multi-agent system in Antigravity
- Models: Gemini 4 Argon, Carbon and Hydrogen
- Math result: Courtade–Kumar conjecture (2013), proved in full
- Formal proof: Lean 4, 50 million lines
- Researcher offer: 10,000 academics, 6 months free, highest usage tier
- Also available: /boost command and Deep Review

## FAQ

### Who can use Antigravity /teamwork today?

Google says external scientists and developers can use Gemini 4 models with /teamwork and /boost in Antigravity today, along with Deep Review. Google Research has also opened applications that give 10,000 academic researchers six months of free access to Antigravity's highest usage tier.

### What is the /boost command in Antigravity?

The Antigravity /boost command packages the iterative coding pattern from /teamwork as a standalone command. Google Research describes /boost as a sequential solve-and-verify loop that was refined from earlier agent trajectories, so developers can run that one pattern without the full multi-agent setup.

### Did the Antigravity agents work alone on these proofs?

Not for most results. Google Research says all of the /teamwork work was done in collaboration with human experts. The exception it highlights is the Weil conjectures formalization, which /teamwork-LEAP produced fully autonomously: about 1.8 million lines of Lean across 28,800 subagent invocations.

### What is Deep Review and who has used it?

Deep Review is Google Research's AI feedback tool for research drafts. Ahead of STOC 2027, authors of 578 submissions, over a third of all submissions, opted in to get Deep Review feedback. In an ongoing survey of 70 authors, 90% said they would use it again. Users get two free Deep Review calls per week.

## Links

- [Google Research Post](https://research.google/blog/advancing-science-engineering-with-antigravitys-teamwork-feature/)
- [Teamwork Launch Post](https://antigravity.google/blog/teamwork-when-ai-becomes-a-research-partner)
- [Deep Review](https://research.google/deepreview/)