platoseedplatoseed
Training Data for Recursive Self-Improvement
We work with frontier AI labs to help train their agents to become AI research scientists
Hillclimb is a virtual lab that trains AI to become researchers by aggregating human research data and automating RL environment creation, aiming for recursive self-improvement. They sell the data they produce to frontier labs and are backed by notable investors and angels.
Hillclimb collects and curates research data to fuel AI-driven experimentation and automatically generates reinforcement learning environments to accelerate AI research. The platform aggregates human research data and automates the creation of RL environments, enabling higher-quality data and scalable experimentation for self-improvement-oriented models. Revenue is generated by selling the data produced to frontier labs.
Who it’s for: Frontier labs and research organizations focused on AI research and model improvement seeking access to training data and automated RL environments.
Hiring for founding engineers; venture-backed by Tier 1 VCs and notable angels
Building the human superintelligence community to advance AI
Hillclimb builds a community of elite math talent to create training data and RL environments for frontier AI labs, starting with math problem-solving data. They claim to partner with researchers to produce data and collaborate with Nous Research, aiming to advance AI training data quality and speed.
Formerly “Plun” · why startups rename →
Training gyms for computer use and software engineering work
Reinforcement learning as a service
