Advancing science & engineering with Antigravity’s /teamwork

Google Research reported on November 13, 2026 that its Antigravity /teamwork multi-agent system, built on the Gemini 4 Argon, Gemini 4 Carbon and Gemini 4 Hydrogen models, produced new results on open problems in information theory, formal verification, molecular structure discovery and kernel optimization, plus several production infrastructure gains at Google. Google Fellow and VP Vahab Mirrokni said the work was done with human experts and that the agentic tools, alongside /boost, are being made available externally in Antigravity, with full features promised to external users in coming weeks. /teamwork addresses two model limits by pairing parallel exploration with dedicated falsifiers such as the Lean 4 proof checker, unmodified software test suites, execution simulators, web search and hardware counters, and by partitioning context so each specialized agent receives only the files and tools it needs.

November 13, 2026 Vahab Mirrokni, Google Fellow and VP, Google Research We share recent results from Antigravity’s /teamwork multiagent systems, which enhance human-AI collaborations to solve open problems across mathematics, chemistry, systems software, and ML engineering. All of this work is done in collaboration with human experts, and we are making these tools available externally. Over the past several weeks, teams across Google and academic partner institutions have tackled open problems in math, science, and complex systems engineering that had remained open for years. Using Gemini 4 Argon https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/ , Gemini 4 Carbon https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-carbon-now-available/ , Gemini 4 Hydrogen https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/ and multi-agent workflows such as Antigravity’s /teamwork https://antigravity.google/blog/teamwork-when-ai-becomes-a-research-partner building on various Gemini models , these collaborations have produced new innovations across information theory, formal verifications, molecular structure discoveries, and kernel optimization, along with several production infrastructure gains at Google. We are making these agentic tools available in Antigravity’s /teamwork alongside /boost https://antigravity.google/docs/boost and other agents so that researchers and developers can apply them to their own work. Frontier research and systems engineering require navigating vast literature, building custom computational pipelines, testing against physical measurements, and verifying every step. Progress comes from an interactive discovery loop where researchers frame questions and define the theoretical or physical constraints, while long-horizon agents write custom solvers, explore large search spaces, test candidate designs, and check long chains of logic. Two limits of current language models motivate /teamwork's design https://antigravity.google/blog/teamwork-when-ai-becomes-a-research-partner/ . First, model generation can be unpredictable, leaving a gap between a single attempt and finding the correct answer. The system bridges this gap by pairing parallel exploration with dedicated falsifiers e.g., the Lean 4 https://lean-lang.org/ proof checker, unmodified software test suites, execution simulators, web search, or hardware counters to automatically verify which candidate solutions work. Second, AI models often lose track of details over multi-day tasks because attention disperses across growing context windows. To prevent this, /teamwork partitions context by construction: each specialized agent receives only the exact files and tools it needs for a specific step. Coordinators hand off tasks to fresh successors, while critics and auditors challenge and verify results from clean contexts without the original producer's reasoning. Teamwork expresses these workflows as declarative self-improving natural-language patterns and dynamically routes each user prompt to the pattern best suited to the task. These patterns range from distributed coding and iterative problem-solving packaged standalone as /boost, a sequential solve-and-verify loop refined from prior trajectories to specialized research workflows like long proof and document review, all of which are available in Antigravity today. This allows teams to design custom specifications from past run transcripts routing across different harness patterns and dynamically improving agentic harness. While many of these features are available in Antigravity, we will provide more updates in the coming weeks providing all the features to our external users on Antigravity. Following our previous posts on Antigravity /teamwork https://antigravity.google/blog/teamwork-when-ai-becomes-a-research-partner/ , we provide new results to open challenges empowered by new variants of /teamwork and Gemini 4 models. Across mathematics and theoretical computer science, researchers paired with Gemini 4 and /teamwork harnesses to collaborate on large-scale proofs and generate complete algorithms and proofs directly from open problem statements. We present several examples. Researchers from Google, CUHK https://www.cuhk.edu.hk/english/research/university-research.html , USC https://www.usc.edu/research-creativity/research-innovation/ , and CMU https://research.cmu.edu/ used /teamwork's Long Proof pattern https://antigravity.google/blog/teamwork-when-ai-becomes-a-research-partner how-the-long-proof-pattern-works implementing Stellar Colosseum https://arxiv.org/abs/2609.15983 in Antigravity to prove https://arxiv.org/abs/2609.24931 information theory's 2013 Courtade–Kumar conjecture https://arxiv.org/abs/1302.2512 in full generality. The result confirms that when a random bit string passes through a noisy channel, no compression of the input to a single bit preserves more information than a "dictator function", which just copies a single input bit. Gemini first found counterexamples to a generalized conjecture from the CUHK team's earlier paper https://arxiv.org/abs/2502.10019 . The CUHK team then proposed a new candidate function that reduced the conjecture to a 14-variable optimization problem, along with a key analytical lower bound. Gemini 4 Argon and earlier Gemini models then partitioned the parameter space and developed most of the new analytical bounds and interval-arithmetic certificates. They also extended the proof from balanced to unbalanced functions without human input. The team formalized the entire proof in Lean 4, producing 50 million lines of code. Researchers also posed open problems directly to Gemini 4 and to multi-agent harnesses such as Stellar Colosseum and /teamwork, receiving end-to-end algorithms and proofs often with Lean 4 formalizations that the authors verified, simplified, and edited for publication. In adaptive data analysis, a multi-agent harness running on Gemini 4 Argon independently proved and Lean-verified near-optimal sample bounds http://github.com/samsonzhou/ada-stream/blob/main/e2e simple.pdf for statistical queries chosen adaptively from earlier answers, closing a gap open since 2014. These bounds yield the first nearly space-optimal algorithm https://arxiv.org/pdf/2611.08699 for data streams with both insertions and deletions that stays accurate against adaptive adversaries. In quantum computing, a new construction settled a 2014 open question https://arxiv.org/abs/2611.09809 by finding the first n -bit Boolean functions needing strictly more than n /2 quantum queries at constant error, while pinning down the exact query overhead across symmetric functions. Several results break long standing barriers in processing and compressing large matrices and graphs. For fifty years until STOC 2024 https://arxiv.org/abs/2311.09095 , combinatorial algorithms for Boolean matrix multiplication and triangle detection https://arxiv.org/abs/2611.06717 avoiding the algebraic cancellations of fast matrix multiplication saved only logarithmic factors over cubic time. A new deterministic algorithm resolves STOC 2024's main open question, achieving its target speedup over cubic time. In high-dimensional geometry, a new lower bound https://github.com/goranzuzic/high-dimensional-steiner-spanners resolves an open problem https://arxiv.org/abs/2608.08706 on sparse networks, proving that approximating all pairwise distances among n points within any constant factor can require polynomially more than n edges, even with extra Steiner points https://en.wikipedia.org/wiki/Steiner tree problem and arbitrarily many hops. In randomized linear algebra, stacked sparse sketches https://arxiv.org/abs/2611.12867 resolve a 2025 open question https://arxiv.org/abs/2508.21189 by breaking the classical sparsity barrier for compressing a matrix to linear dimension without collapsing any direction of an r -dimensional subspace, needing only √log r nonzeros per row quadratically fewer than the log r required to preserve all lengths and nearly matching a new lower bound. For weighted graphs under dynamic edge-weight insertions and deletions, Stellar Colosseum resolved an open question left by FOCS 2022 https://arxiv.org/abs/2209.07729 , developing the first single-pass streaming algorithm for cut sparsification https://arxiv.org/abs/2611.03604 that achieves near-optimal memory, update time, and reconstruction time simultaneously. Our results also span optimization, arithmetic, data structures, and statistics. In optimization, resolving an open question https://arxiv.org/abs/2609.26978 on online inverse linear optimization https://arxiv.org/abs/2610.08735 inferring an expert's hidden objective from decisions over time , the approach gave a deterministic algorithm achieving optimal regret while cutting runtime from exponential in the dimension to polynomial time. For n -bit integer multiplication https://github.com/lockqueue/intmul/tree/main , multi-agent campaigns simplified the theoretical Turing-machine bound below n log n , while a separate effort gave a new O n log n algorithm on a single two-dimensional grid whose C implementation matches the GNU Multiple Precision Arithmetic Library GMP https://en.wikipedia.org/wiki/GNU Multiple Precision Arithmetic Library bit-for-bit up to a billion bits. In data structures, another result made significant progress on Tarjan's 1985 Deque Conjecture https://en.wikipedia.org/wiki/Splay tree Performance theorems whether self-adjusting splay trees handle insertions and deletions at both ends in constant amortized time , proving a new upper bound https://github.com/sorrachai/SplayTree-Deque-Ordinal-Inverse-Ackermann-Bound/blob/main/paper/paper.pdf on the per-operation overhead that simultaneously beats every finitely iterated inverse-Ackermann https://en.wikipedia.org/wiki/Ackermann function Inverse function, verified in Lean 4 https://github.com/sorrachai/SplayTree-Deque-Ordinal-Inverse-Ackermann-Bound . Finally, in high-dimensional statistics, the approach proved https://arxiv.org/abs/2610.27502 that approximate leave-one-out cross-validation https://arxiv.org/abs/1801.10243 , which avoids refitting the model from scratch, remains accurate for non-smooth sparse penalties like elastic net https://en.wikipedia.org/wiki/Elastic net regularization . Checking a computer-assisted manuscript by hand places a heavy burden on peer reviewers, and subtle flaws can hide inside long calculations. Formalizing mathematics in Lean 4 turns a proof into code that a computer kernel checks line by line. Once human mathematicians confirm that the formal theorem statement matches the original problem, the proof only compiles if every logical step holds. Many of the deepest open problems in mathematics govern how physical waves propagate, how dynamical systems evolve, and whether polynomial equations have integer solutions. Working directly from informal proofs, our multi-agent framework, LEAP https://arxiv.org/pdf/2606.03303 , developed with colleagues in Google Cloud AI and paired with Argon, has made progress in these core areas. Notably, /teamwork-LEAP fully autonomously produced an end-to-end, machine-checked formalization of the Weil conjectures https://en.wikipedia.org/wiki/Weil conjectures , the landmark results on counting solutions of polynomial equations over finite fields that culminated in Pierre Deligne https://en.wikipedia.org/wiki/Pierre Deligne 's 1974 proof. This formalization spans about 1.8 million lines of Lean, produced across 28,800 subagent invocations. The system also synthesized over 919,000 lines of Lean to produce the first formalization of a recent proof of Hilbert's tenth problem https://en.wikipedia.org/wiki/Hilbert%27s tenth problem over general rings of integers. It also independently reproduces an end-to-end formalization of a recent proof of finite-time blowup for the 3D Navier–Stokes equations https://en.wikipedia.org/wiki/Navier%E2%80%93Stokes existence and smoothness with smooth forcing, part of one of the seven Millennium Prize Problems https://en.wikipedia.org/wiki/Millennium Prize Problems and open for about 90 years. Beyond these, it also generated the first formalization of the Birkhoff conjecture https://www.unsolvedmath.com/problems/DYN-004 , a dynamical-systems problem open since 1927. Moreover, it formalized for the first time the 3D Bochner–Riesz conjecture https://en.wikipedia.org/wiki/Bochner%E2%80%93Riesz mean Bochner%E2%80%93Riesz conjecture , a central problem in harmonic analysis open for more than 50 years, finding and fixing a gap in a recently released informal proof along the way. Additional machine-checked breakthroughs include resolving the C¹ regularity of infinity-harmonic functions https://www.emergentmind.com/openai-math-explorer/results/377 in three and higher dimensions an open question for about 20 years , formalizing Killip and Vişan's theorem that the KdV water-wave equation is well-posed in H⁻¹ https://annals.math.princeton.edu/wp-content/uploads/annals-v190-n1-p04-s.pdf and verifying a UCLA group's informal proofs of the 40-year-old Jamison Caterpillar Conjecture, the Ağırseven-Ollis conjecture, and an improved upper bound on Erdős Problem 1063. Finally, we provided Deep Review https://research.google/deepreview/ for authors of another CS conference: ahead of STOC 2027 https://acm-stoc.org/stoc2027/ , authors of 578 submissions over a third of all submissions opted in to receive Deep Review feedback on their drafts. In an ongoing author survey N = 70 , 90% of respondents rated the feedback mostly or almost all factual and 90% said they would use the tool again. Over 30% said Deep Review caught substantive proof bugs or mathematically significant typos, and 14% said it found technical gaps that took more than an hour to fix. Surveyed authors shared that “Tools of this kind can help produce cleaner and more precise research, faster” and that DeepReview provides peace of mind on scientific discovery”. We provide users two weekly calls to Deep Review for free. Deep Review https://research.google/deepreview/ is also implemented as /teamwork's document review pattern https://antigravity.google/blog/teamwork-when-ai-becomes-a-research-partner one-framework-many-patterns and available in Antigravity. Read more about DeepReview in our recent blog post https://research.google/blog/analyzing-technical-documents-with-deep-review/ . Beyond formal proofs, pairing specialized generator agents with independent verifier agents enables discovery workflows in experimental laboratory science. Determining a molecule’s structure from experimental nuclear magnetic resonance https://en.wikipedia.org/wiki/Nuclear magnetic resonance spectroscopy data is one of chemistry’s hardest puzzles. For example, in HIV protease inhibitors https://doi.org/10.1126/science.2183354 , one misplaced bond separates an inactive look-alike from a drug. By organizing 30 specialized agents into a self-correcting loop with /teamwork and a physics-informed spectral simulator, we achieved 95% top-1 exact match and 100% top-3 accuracy on complex benchmark molecules, nearly doubling single-agent performance. Beyond teaching Gemini to reason over structure, this lays the groundwork for ongoing works on 3D protein and complex structures dynamics. Beyond this example, science problems including applications in nuclear fusion and quantum algorithms reported in the Gemini 4 blog post https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/ benefited from /teamwork and the Argon model. We also applied a similar multi-round auditing workflow to build empirical benchmarks. Starting from five human-specified acceptance criteria like data anonymization, 29 agents running on Argon over 11.5 hours turned 13 public tables into a 50-task causal inference benchmark. In these tasks, models must determine whether a causal effect https://en.wikipedia.org/wiki/Causal inference can be identified and estimate it against a known answer. After initial checks, separate reviewer and auditor agents caught shortcuts, flagged wording hints, and vetoed hallucinated scripts, turning each into a permanent test. Automating this heavily manual task generation supplies the hard problems needed to help Gemini answer counterfactual "what-if" questions. When paired with objective evaluators such as network benchmarks, game engines, unmodified test suites, and hardware timers, /teamwork campaigns can execute multi-day software and hardware engineering projects. Using /teamwork and Argon, we built FarArray https://antigravity.google/blog/gemini-4-argon-trillion-edge-graphs-and-tpu-kernels , a low-latency kernel-bypass Partitioned Global Address Space https://en.wikipedia.org/wiki/Partitioned global address space based on the SNAP microkernel https://research.google/pubs/snap-a-microkernel-approach-to-host-networking/ . Applications built on FarArray include lock-free Connected Components https://github.com/ParAlg/gbbs , which scale to 1-trillion-edge graphs with 60x to 90x throughput gains and 40x to 180x speedups at 6x lower RAM over TCP key-value stores. Similarly, InstantEmbedding https://arxiv.org/abs/2010.06992 Personalized PageRank reduced peak DRAM by up to 14x and runtime by 11x to 27x over accelerated MapReduce implementations. The team also used /teamwork and Argon to create custom TPU kernels for Graph Neural Networks, delivering up to a 185x High Bandwidth Memory https://docs.cloud.google.com/tpu/docs/system-architecture-tpu-vm how a tpu works reduction and 8x to 15x speedups in forward and backward passes. Further details are available in our recent Antigravity blog post https://antigravity.google/blog/gemini-4-argon-trillion-edge-graphs-and-tpu-kernels/ . To test autonomous optimization over long horizons, hill-climbing teams on Argon trained a 9x9 Go engine https://en.wikipedia.org/wiki/Computer Go from scratch starting from a basic AlphaZero https://github.com/michaelnny/alpha zero implementation. Human researchers set the chip allocation and evaluation setup, while the harness ran autonomous campaigns for faster self-play achieving a 2,300x speedup and faster Elo https://en.wikipedia.org/wiki/Elo rating system readings, promoting champions strictly via statistical tests on locked records. In under eight days, the engine gained roughly 2,600 Elo, reaching 1,885 Elo, placing it between KataGo https://github.com/lightvector/katago at 16 and 32 visits per move. Large code migrations fail when agents stub out dependencies or alter tests to make a broken port pass. A new /teamwork rewrite pattern, soon to be available in Antigravity, solved this by scheduling each module only after its dependencies compile, linking every leaf module against the original, unmodified test suite. Building on the initial 25% port of the Zircon microkernel https://en.wikipedia.org/wiki/Fuchsia %28operating system%29 , a two-week Argon campaign rewrote the remaining C++ code, adding about 800,000 lines of safe Rust that pass all kernel unit tests and boot on arm64 hardware. /teamwork also rewrote libgav1 https://chromium.googlesource.com/codecs/libgav1/ , Google's AV1 video decoder, from C++ into around 170k lines Rust, achieving 100% bit-exact parity across a broad conformance test suite. Initially, the port was around 4x slower than the original C++ code. However, after giving /teamwork access to profiling tools and a mandate to maximize speed on sample video clips, the Rust port became 20-30% faster than the original C++ and hand-optimized assembly code on both x86 64 and ARM. For day-to-day coding tasks, the Antigravity /boost https://antigravity.google/docs/boost command packages /teamwork's iterative coding pattern https://antigravity.google/blog/teamwork-when-ai-becomes-a-research-partner one-framework-many-patterns . A long-running /teamwork campaign analyzed past run trajectories and found that tightly coupled code is best handled through sequential implementation and revision stages rather than parallel splits, improving /boost on coding tasks. This is an example of a harness optimizing another harness for token efficiency. For machine learning systems, full hardware evaluations of custom JAX and Pallas TPU kernels take 7 to 32 minutes, limiting how many candidates prior optimization agents can explore. Using /teamwork's DuoDiscover https://arxiv.org/abs/2610.17184 pattern on Argon, four parallel agents identified performance bottlenecks and constructed fast, 23-second proxy evaluators called yardsticks. These yardsticks produce candidate orderings that are continuously validated against periodic full hardware runs using Kendall's rank correlation https://en.wikipedia.org/wiki/Kendall rank correlation coefficient . DuoDiscover achieves a 4.09x geometric-mean speedup over reference implementations across 17 JAXBench kernels and a 2.99x geometric-mean speedup across 7 hand-optimized vLLM TPU-Inference Pallas kernels for frontier open-weight models. We have also applied /teamwork and /boost to many production infrastructure and hardware modeling problems across Google. Examples include: External scientists and developers can use Gemini 4 models with /teamwork and /boost in Antigravity and Deep Review https://research.google/deepreview/ today. We have also opened applications https://forms.gle/WvmBbPvRbpKG1BPn9 giving 10,000 academic researchers six months of free access to our highest usage tier. Upcoming updates will bring our formal proof pattern and additional specialized workflows into Antigravity. These features will be accompanied by deep-dive tutorials on the Antigravity blog https://antigravity.google/blog in the coming weeks, ensuring that developers and researchers have the resources needed to generate, inspect, and iteratively refine custom multi-agent teams. We are excited to make our tools in addition to our models available for everyone. Antigravity’s /teamwork feature is a result of a Google-wide effort with many contributors and ongoing explorations. Contributions include: Yuan Deng , Honghao Lin , Jieming Mao , Vahab Mirrokni , Vishnu Veeramachaneni , David Woodruff , Song Zuo , Shubham Agarwal, Santiago Balseiro, MohammadHossein Bateni, Mohsen Bayati, Garrett Bingham, Guy Blelloch, Douglas Chen, Vincent Cohen-Addad, Piaoyang Cui, Tulsee Doshi, Eric First, Michael Galkin, Elena Gribelyuk, Anupam Gupta, Guru Guruganesh, Mike Heddes, Patrick Heisel, Iman Hosseini, Kevin Hou, Lalit Jain, Adel Javanmard, Rajesh Jayaram, Xiaoen Ju, Po-Nien Kung, Silvio Lattanzi, Samuel Lavoie, Chun-Liang Li, Matthew Li, Yinan Li, Yuchen Li, Yiming Liu, Thang Luong, Brandon Mayer, Simon Meierhans, Swaroop Mishra, Ashish Naik, Sebastian Nowozin, Renato Paes Leme, Nikos Parotsidis, Ali Parviz, Neel Patel, Violet Peng, Bryan Perozzi, Anshul Ramachandran, Shashir Reddy, Lijie Ren, Madhavi Sewak, D Shin, Reza Shokri, Sukhdeep Singh, Samuel Smith, Linfeng Song, Dilara Soylu, Zheng Tan, Sasan Tavakkol, Hadi Tavassol, Yifeng Teng, Tom Wollschläger, Haifeng Xu, Zicheng Xu, Yihao Xue, Amir Yazdanbakhsh, and Goran Žužić. Carla Bromberg and Lauren Fraser for program management, and Francis Duan, Pan-Pan Jiang, Xiyan Shao, Jane Shapiro, Weilun Shi, and Xinyu Ye for their contributions to the project. We also thank support from several leaders Sergey Brin, Demis Hassabis, Koray Kavukcuoglu, Pushmeet Kohli, James Manyika, Yossi Matias, Varun Mohan, and Amin Vahdat. Several academic collaborators contributing to the research papers: Zijie Chen, Amin Gohari, Chandra Nair CUHK , Songbin Liu, Arian Maleki, Chen Shang, Hanbo Yang, Wenjin Zhang Columbia University , Zihan Hao, Qipeng Liu, Jianan Shao UCSD , Lorenzo Beretta IBM Cambridge , Simone Bombari BYU & Flatiron Institute , Mahbod Majid, Margalit Glasgow MIT , Raphael Arkady Meyer UC Berkeley, ICSI, & LBNL , Kijun Shin CMU , Huacheng Yu, Elena Gribelyuk Princeton University , Xin Lyu UCB/ ByteDance , Samson Zhou Texas A&M , Erik Waingarten UPenn , and Sorrachai Yingchareonthawornchai ETH-ITS .