Local models
Every patch has a view.
Bounded, self-reading software patches hold local state and records. Ports determine which sensors and other models each patch can read.
Pragma Research / Investors
Build the brains for the physical world.
Cadence is a deep real-time brain that learns from experience. Our ambition is to put it inside robots that keep improving throughout their working lives.
Proposed seed round: $4 million over 18 months.
The company thesis
Pragma Research builds the brains. Robot manufacturers build the bodies.
We want a robot's experience to become part of its intelligence.
A machine encounters changing loads, unfamiliar conditions and the consequences of its own actions. Cadence is built around a persistent network that participates in that loop and learns from it.
Robotics is our first market. We propose a paid evaluation on one valuable task, followed by deployment of the same brain architecture across more tasks and bodies.
Explore the robotics opportunity →The discovery
A theory of cooperating observers, made executable.
Cadence brings together observer-based physics and ideas from cortical organization. Local predictive models settle to produce an answer, and learning repairs their retained relations through experience.
Three supported layouts
Connect the models the task needs.
Flat layouts read sensors directly. Ordinary state-coupled layouts add intermediate representations. Recursive observers also read exact prediction errors. All three share the same settlement and qualification procedure.
Experience
The same system learns.
Experience changes the brain's retained relations. Public evidence bundles connect specific learning results to the code and experiments that produced them.
The software tests a computational hypothesis inspired by the cortex. Biological identification is a separate scientific question.
Evidence in simulation
A working core. A measurable learning result.
In a reserved Doom Basic evaluation, the selected learner succeeded in all 64 episodes, compared with 53 for the original founder baseline. The candidate combined further experience with a selected replay-setting change.
The task uses finite teaching, simulator-assisted practice and replay. The result belongs to this scenario and evaluated candidate.
Read the evidence and conditions →64/64
Selected learner
Doom Basic successes
53/64
Original founder baseline
Same reserved evaluation
The business
One integration can become a fleet business.
The proposed product is a supported Cadence runtime and a brain prepared for the customer's embodiment. The commercial relationship begins with a concrete evaluation.
01
Paid evaluation
Integrate one task with a robot maker. Measure the value of adaptation against its existing stack and agreed resource budgets.
02
Production license
Convert useful results into a runtime and embodiment license, with deployment integration and support.
03
More deployed brains
Expand through additional robots and tasks. Reusable software and integration tools are the route to recurring fleet revenue.
Proposed seed round
$4 million. 18 months. A physical and commercial leap.
We are seeking a lead investor and participants to help turn the working software into a deployable robot brain. The proposed programme targets three outcomes.
01
Physical proof
Demonstrate useful adaptation on one body, retaining learned behavior within measured control and compute budgets.
02
Customer value
Complete a paid evaluation where learning reduces intervention or increases productive operation.
03
Repeatable integration
Transfer the software to a second task or integration and establish a path to a deployment license.
The financing proposal pairs an 18-month programme with a $4 million ask. Staffing, hardware access and milestone budgets form part of investment diligence.
Founder
Bernhard Mueller
A career in reverse engineering, applied to the mechanisms of physics and intelligence.
Bernhard leads Pragma Research's research and the development of Cadence. The work connects a mathematical research programme with executable software and inspectable experimental evidence.
The proposed round adds robotics integration and systems engineering capacity alongside learning research. We welcome investors who want to engage with the architecture and help build the company around it.
Request a founder demo →