Thesis is an applied AI research lab automating AI research and development. The company aims to help researchers discover breakthroughs like AlphaFold or the Transformer faster, more systematically and at scale by treating scientific discovery as a combinatorial search problem. Thesis has reported state-of-the-art results on OpenAI's MLE-Bench, which tests how well AI systems can autonomously train machine learning models. It is based in San Francisco and is part of Y Combinator's Fall 2025 batch.