ExperBrain Open-Source Project
Africa's Open-Source Path to AGI

Allan Ebusuru

From the moment I first read about predictive processing and the idea that the brain is a prediction engine, something clicked. I spent the next three years reading, simulating, and building — not in a corporate lab, but in one room in Nairobi. I was driven by a single conviction: the path to robust, grounded intelligence requires an architectural rethink, not only more scale. That conviction became the Cortical Language Model (CLM). CLM is a first-principles architecture inspired by predictive coding, sparse modular computation, and embodied cognition. It replaces monolithic, dense networks with a dynamic routing core (the “thalamus”), specialized expert modules (cortical columns), and a multi-modal grounding pipeline that ties symbols to sensory experience. The whitepaper is the technical record of that work. The codebase contains the early reference implementations and small-scale simulations that demonstrate the core ideas: a dynamic router that learns sparse expert activation, per-expert working memories, and a predictive objective that trains the model to minimize future sensory surprise rather than only predict next tokens. I founded ExperBrain to make this idea collective. The point isn’t to build a proprietary product — it’s to provide an open architecture and invite neuroscientists, ML engineers, data curators, and ethicists to help prove (or disprove) whether CLM is a viable path toward AGI. If the architecture is sound, it should stand up to open scrutiny and grow faster with more minds pushing on it.