Autopoiesis Sciences

We are building AI that can do science autonomously.

Science is the hardest thing humans do. It requires sustaining inquiry under uncertainty, following weak signal through long silences, and recognizing that absence of result is not the end of a problem. It is often the beginning of the right one. Every major discovery was made by a mind that refused to stop. Current AI systems do not know this. They exit when tools go quiet. We are working on changing that.

We believe the bottleneck to scientific discovery is not data or compute. It is judgment. Knowing which directions are worth pursuing before the evidence arrives. We are building the architectures, training methods, and reasoning systems that give AI that judgment.

We work on the full stack: from how scientific reasoning is distilled into models, to how agents navigate long-horizon problems with sparse reward, to how discovery itself can be made autonomous. In the near term, Aristotle, our current system, gives scientists a powerful research partner. In the long term, we remove the need for one.

We are backed by Informed Ventures, Alpaca VC, and VitalStage Ventures, with participation from Adam Grosser, Ally Warson, Mike Mahklow, Hiro Mizushima, Tomas Urena Munoz, Vanessa Cann, and Cross Atlantic Angels. For more information, get in touch: contact@autopoiesis.science