Key Findings
Google has expanded its Antigravity multi-agent system to solve complex mathematical problems and optimize software systems. Most notably, the Teamwork-LEAP system generated approximately 1.8 million lines of Lean code across 28,800 subagent invocations to formalize the Weil conjectures, originally proved by Pierre Deligne in 1974. Additionally, the system reproduced a formalization of a proof regarding finite-time blowup in three-dimensional Navier–Stokes equations, a problem that has remained open for about 90 years.
Technical Details
The Teamwork framework employs multiple specialized agents working in parallel to reduce reasoning errors. Each agent is restricted to the specific files and tools required for its task, while independent critics and auditors verify results without relying on the original agent’s reasoning. This approach was applied across three primary domains:
- Mathematics: Researchers used Lean 4 to formalize the Courtade–Kumar conjecture, with Gemini models deriving analytical bounds and generating computer-checkable certificates.
- Chemistry: A team of 30 specialized agents and a physics-informed simulator identified molecular structures from NMR data with 95% top-one accuracy and 100% top-three accuracy.
- Software Engineering: Agents rewrote the libgav1 AV1 video decoder from C++ into approximately 170,000 lines of Rust. After using profiling tools for optimization, the Rust version performed 20% to 30% faster than the original C++ implementation on ARM and x86_64 processors.
Background & Context
Beyond the AV1 decoder, Google used the system to rewrite the Zircon microkernel’s remaining C++ code into approximately 800,000 lines of memory-safe Rust over a two-week campaign. In internal infrastructure, an agentic scheduling workflow reduced TPU workload preemption by 24% on certain hardware types.
Significance & Outlook
Google is making the Teamwork framework and its /boost iterative coding workflow available through Antigravity. The company has opened applications for 10,000 academic researchers to receive six months of free access to its highest usage tier, with plans to release additional specialized formal proof tools in the future.
Source: https://interestingengineering.com/ai-robotics/google-antigravity-multi-agent-research
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime



