When information takes shape
Our research perspective on how data can form a field, how reasoning can build within it, and what would be needed to test the idea.
ResearchField-based intelligence
Can intelligence compound?
Mumbrane researches field-based intelligence: AI models that combine general language capabilities with bounded data and user-defined reasoning.
01Three connected lines of inquiry
The questions behind the model.
Our central question is what happens when a result becomes the starting point for further reasoning. Can that process extend what a model can establish from the information it has? What makes it useful, and what limits it?
01Representation
How should information form a field?
How representing data and reasoning together shapes a model’s conclusions.
02Dynamics
When does one result help the next?
When intermediate results support further reasoning or carry errors forward.
03Causality
What sets the boundary?
How data and reasoning set a model’s limits, and how those limits change.
02Inspired by general relativity
Relativity as a starting idea.
Our inspiration comes from general relativity, where matter and energy influence spacetime geometry. We explore a computational analogy: data shapes the field within which a model reasons.
03From hypothesis to experiment
Testing the idea through Paladin.
Paladin lets us test field-based intelligence on concrete tasks. We assess whether compounding improves results, how errors carry forward, and how changes to data or reasoning affect answers.
Research perspectives
04Work with the lab
Bring a question to the lab.
We welcome research questions, proposed experiments, and collaboration inquiries on field-based intelligence, compounding reasoning, and model evaluation.
Contact the research team