Loading organizations...
Loading organizations...
hillclimb: Training Data for Recursive Self-Improvement
Based in San Francisco, California, Hillclimb is an early-stage artificial intelligence startup that provides specialized training data for recursive self-improvement to help develop autonomous AI research scientists. The business-to-business company partners directly with frontier artificial intelligence laboratories, supplying advanced reinforcement learning environments and datasets that enable AI agents to iteratively enhance their own complex research capabilities. Operating with a lean team of four employees, the enterprise is backed by Y Combinator and was selected to participate in the prominent accelerator's Fall 2025 startup batch. Previously operating under the corporate name Plun, the firm focuses specifically on the intersection of software engineering, product design, and scalable machine learning infrastructure to support next-generation model development. Hillclimb was originally established to provide foundational data infrastructure for highly automated machine learning research workflows across the broader technology sector.