Research
We explore the structure of machine cognition. Understanding how neural networks think, reason, and represent knowledge.
Research Areas
Mechanistic Interpretability
Understanding neural networks by reverse-engineering their internal computations. Finding the circuits that implement specific behaviors.
Sparse Autoencoders
Training networks to decompose neural activations into interpretable features. Making the latent space legible.
Model Reasoning Inspection
MRI: A framework for understanding how language models reason. Tracing the path from input to output through interpretable steps.
Neural Topology
Studying the geometric and topological properties of neural network representations. The shape of thought in latent space.
The Three-Phase Model
Our approach to understanding neural network reasoning
Activation Capture
Recording neural activations across model layers during inference.
Feature Extraction
Decomposing activations into interpretable features using sparse autoencoders.
Reasoning Analysis
Tracing feature activations to understand the model's reasoning process.
Publications & Datasets
Neural Topology and the Geometry of Thought
Exploring how information flows through neural networks by analyzing their topological structure. Understanding the shape of machine cognition.
Sparse Autoencoder Feature Dictionaries
Pre-trained SAE feature dictionaries for common language models. Enabling interpretability research without the computational overhead of training from scratch.
MojoMem: Adaptive Memory Systems
A patented approach to adaptive memory management in AI systems. Enabling more efficient and contextually-aware information retrieval.
Nomos Logos
Our AI research platform. A unified environment for interpretability research, model analysis, and reasoning inspection.
Currently in development. Join the waitlist for early access.
Request Access