AMEN / Music
Make a jungle dub.
A tiny trained composer generates music from its learned record memory. Change the variation and listen.
Pragma Research / Machine intelligence
Brains that learn.
Cadence is a deep real-time brain that learns from experience.
We are building the intelligence for robots that keep improving throughout their working lives.
The Cadence core is open source, with documentation and runnable examples. Anyone can build on it.
Cadence / Settling networkCoupled models
Architecture illustration
The idea
Small models.
A shared answer.
Cadence connects local predictive patches. Each holds state and records, reads through its connections, and repairs prediction mismatches.
Flat settlement learns direct responses. State-coupled settlement coordinates populations. Recursive observers also read the brain's own prediction errors.
Working demonstrations
Try a learning brain.
Patch World / Artificial life
Watch a world learn.
Creatures forage, learn and evolve their brain layouts. Select one to inspect what it sees and believes.
Learning you can measure.
In one reserved Doom Basic evaluation, a selected learner achieved 64 successes in 64 episodes, compared with 53 for its original founder baseline. Training combined simulator-assisted experience with a changed replay setting; this result covers a defined simulation task.
Build with us
We build the brains.
You build the bodies.
For robotics teams.
Bring one body and a measurable adaptation problem. Let's test what a learning brain changes.
Explore a partnership ↗For investors.
Help take a working software brain from simulation into physical robots.
The investment case ↗Physics, mathematics and AI
Research behind
the brain.
Pragma Research is led by Bernhard Mueller. Cadence grew from Observer Patch Holography: research into how local observers build a shared account of a world.