Body
Changed response
A different payload or a worn motor changes what a command does.
01 / 08 · Pragma Research
Real-time AI that learns from live experience.
Cadence is modeled after animal and human brains. We build the brains for robots that keep improving throughout their working lives.
Proposed seed round
$4 million over 18 months
From a working software brain to physical performance, a paid customer evaluation and repeatable integration.
Bernhard Mueller · bernhard@floatingpragma.ai · floatingpragma.io
02 / 08 · The customer opportunity
A robot's experience should make it more useful.
A body changes. A familiar movement produces an unfamiliar result. The commercial question is how quickly the robot can recover.
Operation
Recurring intervention
The same failure draws an operator away from other work.
Value
Learning that pays
Our first evaluation targets less intervention or more productive operation.
First market: robot makers and integrators with one reproducible adaptation or recovery problem.
03 / 08 · The technology
Modeled after animal and human brains.
Most of today's AI runs on transformers: completely feed-forward, and typically deployed with attention matrices and all other weights fixed once training ends. Cadence runs in real time and keeps learning from live experience.
It learns on the job
The same brain acts and learns from the result of every action. There is no training phase before deployment.
It adapts to change
A worn motor or a new load shows up in what the robot senses. The brain adjusts without being told what changed.
It does not act on a guess
Cadence acts only on a settled answer. When it has none, it reports that and the robot can fall back.
Cadence is built from small self-reading patches, each with its own state, connections, records and feedback. They settle on one answer together. How Cadence works
04 / 08 · What exists today
A working brain you can watch learn.
Atari Arcade
Two untrained brains start when the page opens. They watch a teacher and take over the controls within minutes.
Patch World
Creatures learn within one life, and evolution reshapes their brains across generations.
Rover Lab
A simulated rover's wheel is weakened without warning. The live brain relearns its body and keeps reaching its targets.
The software runs today and its core is open source. The funded programme takes the brain out of simulation and into a physical robot.
Watch the demos →05 / 08 · First commercial engagement
One body. One task. A paid evaluation.
We are seeking a robot maker with a concrete adaptation problem and an engineering sponsor.
Agree the test
Choose the task, what the robot senses and what experience it gets. Define success.
Integrate Cadence
Connect the brain to the robot or simulator. Compare with the customer's existing stack.
Make the deployment decision
Measure time to recover, operator interventions, response time and running cost.
Proposed physical demonstration: adaptation to changed motion response on a small rover or tabletop robot. Robotics evaluation programme
06 / 08 · The business model
Pragma Research builds the brains. Manufacturers build the bodies.
A reusable intelligence product, licensed across tasks and robot fleets.
01
Paid evaluation
A scoped integration and a measured answer to a customer problem.
02
Production license
A supported Cadence runtime, an embodiment package and deployment integration.
03
Fleet expansion
Recurring licenses for more robots and tasks, supported by reusable tools and deployment knowledge.
Proposed commercial model around the GPL-3.0 core. Licensing arrangements form part of product development. A second integration tests reuse.
07 / 08 · Founder and company
From reverse engineering to a working brain.
Bernhard Mueller / Founder
Creator of Mythril and a creator and lead author of the OWASP mobile security standards and testing guide.
Bernhard leads the research and development of Cadence. The work spans an open-source core, preprints and live demonstrations.
What the round adds
A team for physical deployment.
Robotics integration, systems engineering and learning research turn the core into a product a manufacturer can evaluate and deploy.
We welcome technical investors who can help recruit the team and connect it with robot manufacturers.
08 / 08 · Proposed seed round
$4 million over 18 months.
Help take a brain that learns from live experience out of simulation and into robots.
Physical proof
A real robot adapting to a change that defeats a frozen brain.
Customer value
A paid evaluation that connects learning to productive operation.
Repeatable integration
A second task or integration and a path to a deployment license.
Proposed programme and financing target. Staffing, hardware access and milestone budgets form part of investment diligence.
Seeking a lead investor and round participants.
bernhard@floatingpragma.ai · floatingpragma.io/investors