AI Research

Machines that teach themselves

We're building the foundations for long-horizon learning and recursive self-improvement. Currently in stealth.

Our research Get in touch
Backed by
To be announced
With angels and advisors from
Handshake.ai Amazon AWS Carnegie Mellon University
What we work on

Core research areas

We pursue fundamental questions about how intelligent systems can continuously improve through interaction with their environment.

01
Long-Horizon Learning
Training agents that develop capabilities over extended interaction sequences, building compounding understanding through practice and feedback.
02
Recursive Self-Improvement
Systems that identify their own weaknesses, generate targeted training, and iteratively enhance their capabilities without human intervention.
03
Robust Evaluation
Ensuring that measured progress reflects genuine capability gains, not artifacts of the evaluation process itself.
Our Belief

Models that teach themselves through practice.

Join us

We're hiring

We're a small team working on hard problems. If you want to shape how AI systems learn to improve themselves, we'd like to hear from you.

Research Scientist
Full-time · Remote

Let's talk

We're in stealth. If our research interests you, reach out.

View open roles