What the loop actually does
Every iteration of Axiom's autonomous loop has the same shape: observe the platform state, identify gaps + risks, classify impact, pick the highest-leverage safe action, check policy + security boundaries, dispatch to a typed runner, validate the output, write a trace + audit + memory event, then continue. The dispatch step is intentionally paused — the brain never runs a destructive action without explicit operator approval.
Hard safety contract
The loop refuses to escalate. No destructive autonomy. No secret exposure. No silent execution. No fake live state. Every audit event records the policy decision and the readiness factors that allowed the step. Preview-mode sources are physically prevented from reaching execution-only stages.
Composition over chat
The brain doesn't generate Terraform from a prompt. It composes typed signals — capability coverage map, validation report, command center state, security reasoning, release readiness, planning candidates — into a structured BrainOperationsResult with explicit evidence references. No hidden chain-of-thought; every conclusion pins to a source id.
What this unlocks
Trustable automation. The operator can ask 'why' and get a citation, not a hallucination. The auditor can replay every decision against the same typed inputs. The brain knows what it can and cannot do — the capability coverage map is the single source of truth for which actions are live, preview, or planned across every provider.