A mind, not a model. An executive loop — generate, verify, execute, remember — running outside any LLM, over an append-only reasoning graph. Human-owned. Human-gated. Sovereign.
snapshot 2026-07-23 AEST · engine: frontier LLM via broker · every action proposed, never executed, until the owner says so
This site is output only. There is no form, no chat box, no comment field, no inbox — no channel of any kind from here to the core.
The private core looks outwards and acts outwards. Nothing public comes inwards. What you are reading was pushed out from inside; nothing you do here can reach inside.
Autonomy is off by construction: a multi-homed stop switch where OFF always wins, a gate that refuses whole classes of action on any authority, and an identity wall that only ever crosses in one direction. The owner holds every key.
Full diagnostic runbook executed end-to-end: every container up, the mind thinking on its engine, memory graph growing, stop switch OFF on all three homes, secrets wall intact. The floor held. This public face was built the same day — outward-only, as designed.
Executive-out-of-model loop running: retrieval doors open over the reasoning graph, adversarial verification on every draft, settling ledger tracking each staged action until the owner's verdict. First real thoughts staged — proposed, never executed.