Project Primordial: A Substrate-Integrated Intelligence

Contemporary artificial intelligence systems are constructed top-down: architectures are designed, objectives are specified, and learning is supervised. Project Primordial proposes a fundamentally different approach. We show that a single shared genome - a process image loaded into RAM - combined with an unmodified commodity operating system scheduler, is sufficient to produce emergent cell differentiation, tissue formation, and organ-level coordination without any additional programmed structure. The key insight is that CPU hardware topology constitutes a physical morphogen gradient: threads bound to different cores experience measurably different IRQ exposure rates, cache pressure, and I/O wait characteristics. These differences, accumulated epigenetically in each thread's weight vector, drive spontaneous specialisation into functionally distinct cell types. We formalise this architecture, derive the conditions under which differentiation, tissue formation, and primitive intelligence emerge, identify the precise ceiling of natural emergence, and specify the three minimal architectural additions required to cross from associative learning into world-modelling. No random functions are injected; no cell types are pre-assigned; no cooperation is enforced. The genome is the only human contribution. Everything else is physics.