An energy-based engine that learns systems while they run.

Most machine learning is fitted once and then used. The systems people run keep changing.

Demand shifts. Links between parts appear and break. Some rules always hold. Eyrie's engine treats a running operation as one coupled field. It learns from every event as it becomes known, never from hindsight, and says how sure it is. It is built to tell when the structure of a system has changed.

It is not a language model. It learns from what a system does, not from text about it.

• 01Prediction to decision, in real time. For operations where each decision changes what happens next: weighing a decision by playing the system forward before it is made. Healthcare operations and bioprocess.
• 02Learning and managing systems. For large systems run by people and AI agents: learning how they actually behave, and judging whether that behaviour, including the work of AI agents, still matches what was intended.

Each claim is written down and dated before its outcome is seen, and published whether it passes or fails.

Sarah Todd Davidson, Founder
Independent researcher and founder. Builds energy-based learning engines from physics and algebraic topology, and puts them to work on real operations. Also founder and CEO of PAUL Health. sarah-in-dev.com
sarah@eyrie.systems