Research
Dynamics, geometry, memory, interaction.
We study intelligence as an emergent consequence of structured dynamics and interaction, with an emphasis on mathematical mechanisms that improve representation, adaptation, reasoning, and generalisation.
Directions
- 01
Dynamics of Intelligence
Recurrent computation, structured memory, temporal dynamics, reservoir computing, and emergent representations for adaptive intelligent systems.
- 02
Geometry & Structure in Learning
Hyperbolic representations, topology, equivariance, symmetry, and geometric inductive biases for representation and generalisation.
- 03
Agentic & Adaptive Systems
Memory, interaction, reasoning, and dynamical mechanisms underlying autonomous and continually adapting machine intelligence.
Philosophy
Rather than treating time, memory, geometry, or biological structure as afterthoughts, we ask when they should be first-class computational primitives. The goal is not simply to make models larger, but to make their inductive structure better aligned with the processes they must represent.
Inspirations
Continuous Thought Machines
A dynamics-first architecture where neural timing and synchronization act as core computational primitives.
Baby Dragon Hatchling / The Dragon Hatchling
Brain-like scale-free interaction graphs, synaptic plasticity, and interpretability as a bridge toward biologically grounded computation.