Your browser does not support the video tag. A year ago, we unveiled AlphaEvolve, a Gemini-powered coding agent designed to create cutting-edge algorithms. We demonstrated that AlphaEvolve can drive new discoveries in longstanding open problems in mathematics and computer science, while also optimizing algorithms now running in key parts of Google’s infrastructure. Today, with algorithms embedded in virtually every domain of life, the range of possibilities for AlphaEvolve has grown even wider. From advancing our understanding of the natural world’s physics to optimizing electricity grids and computing systems, AlphaEvolve offers countless ways to accelerate progress for scientists and businesses across diverse fields. We’re excited to highlight some of its most significant achievements so far, including driving meaningful social impact and advancing sustainability. AlphaEvolve has revealed important links in health and sustainability research, pushing the boundaries of scientific discovery. AlphaEvolve is acting as a powerful research partner, speeding up scientific discovery across multiple fields. Moreover, its ability to drive autonomous discovery is sparking parallel breakthroughs in a wide range of other domains — from uncovering interpretable neuroscience models and establishing new theoretical limits in microeconomics, to rapidly improving neural network components, privacy-preserving cryptography, synthetic data generation, and essential safety measures for cutting-edge AI systems. AlphaEvolve optimizing an instance of the Tammes problem. You can browse a variety of other problems for which AlphaEvolve produced candidate solutions in the public Gallery. Improving AI infrastructure. AlphaEvolve has moved beyond pilot testing and is now a core part of our infrastructure.
Google DeepMind News