Hillclimb

Hillclimb aims to improve model capabilities toward recursive self-improvement. To do so, the company focuses on solving two core problems: aggregating all human research data and automating RL environment creation. Each step forward on these axes unlocks further data, enabling stronger research capabilities to emerge, bringing the world closer to self-improving models. Hillclimb is backed by Tier 1 VCs, Paul Graham, and angels from OpenAI, Anthropic, DeepMind, xAI, and Meta Superintelligence Labs.