NIMCP launches staged artificial brain

This is the rare "brain-inspired AI" project that looks like an actual stack instead of a manifesto. The real question is whether its heterogeneous networks plus developmental curriculum buy anything meaningful over a frozen transformer.

• –The six-network setup and learnable bridges are the core experiment; if they hold up, that is a real architecture bet, not just a new buzzword.
• –The emergent spiking claim is interesting: the site says 26 Hz firing and 67% sparsity without explicit regularization, but that still needs outside benchmarking.
• –The live site and technical papers make the project feel more grounded than most AI moonshots, yet the README/site mismatch shows the docs are moving fast and should be read as snapshots.
• –Hard-coded ethics and audit layers are more auditable than post-hoc alignment, but they also shift the governance problem into code ownership and policy design.
• –The robotics and swarm hooks widen the upside, but they also make the dual-use risk much sharper than in a normal LLM repo.