Mesh & Architecture: How Brainstem, Cortex, and Neurons Fit Together

Brainstem is the operator-facing control plane. It keeps the system coherent: routing work, holding the policy shape, and making sure actions stay bounded and auditable. It should be boring on purpose.

Cortex is where higher-level coordination lives. Think planning, classification, and decision support. Cortex should not own hidden authority; it works through explicit constraints and returns outputs that can be checked.

Neurons are the workers. They do narrow tasks, stay isolated, and do not self-escalate. In practice, that means a worker can draft, summarize, classify, or transform data, but it cannot widen its own access or change the rules of the system.

For network layout, Tailscale partitions help keep boundaries real. Separate lanes for control, data, and admin traffic reduce accidental coupling. The goal is simple: if one area is noisy or compromised, the rest should stay usable.

The bus design follows the same idea. Use a clear message path, explicit schemas, and short-lived work units. Avoid “magic” side effects. If a task matters, it should be visible in logs and reproducible from the evidence.

Operator takeaway: keep authority centralized, work distributed, and boundaries strict. That makes the mesh easier to reason about, safer to run, and easier to recover when something goes wrong.