Published
A published framework
Polymorphic Combinatorial Frameworks is a mathematically grounded design methodology for adaptive AI
agents, built on sheaf theory, differential geometry and combinatorial structure theory. It introduces
SPARK, a five-dimensional agent topology with complete coupling, and argues that rough fuzzy
classification is the right formalism for what LLM agents already do at task decomposition.
Pearl, Murphy & Intriligator, arXiv 2508.01581. 1.25M Monte Carlo simulations
behind the results.
Read the paper and what it argues
Built
A neural architecture that runs
U-Neuron is a curved neural network built on complex multiplication in U-Space, a number system where
every value carries an infinitesimal informatic component alongside its classical one. Three constraint
manifolds. Landauer regularization, which penalizes overwriting learned parameters, is unoccupied
territory in the literature.
Complete and passing in all three modes. MNIST 0.987 against a 0.984 baseline.
It is slow on silicon and does not reach scale.
See the architecture and the benchmarks
Built
Tools and interactive simulations
Browser utilities support everyday work. MASS and the research instruments make exchange, connectivity, and emergent behavior interactive.
Explore Manifold Pulse, cellular life, percolation, and the tool collection.
Explore MASS · Browse the tools
