What an LLM cannot reliably prove, verify, replay, validate, source, audit, or certify, GoalOS converts into AGI Jobs.
LLM gives candidate output → GoalOS checks what is unsupported → unsupported claims become Proof Debt → Proof Debt becomes AGI Jobs → AGI Agents / humans complete the jobs → AGI Nodes / reviewers validate them → ProofBundles return → Evidence Docket updates → only validated capability can compound.
Paste an LLM answer. GoalOS will treat it as a candidate draft, not authority. Claims that lack proof become proof debt and then AGI Jobs.
The authority layer is the gate. LLM output can draft, suggest, and reason; it cannot promote itself.
Unvalidated candidate output cannot enter Chronicle memory or influence future missions.
GoalOS marks which draft claims are supported, weak, unsupported, or blocked.
| Claim | Verdict | Why | Required proof |
|---|
Readiness rises only when evidence, replay, validator verdicts, and Evidence Docket updates are present — not because the LLM sounds confident.
Proof debt becomes executable work. Each AGI Job includes worker route, validator route, required ProofBundle, acceptance tests, and blocked claims.
| Job | Proof Debt | Worker | Status | Actions |
|---|
Paste a ProofBundle or use generated job returns. Promotion remains blocked until replay and validator verdict pass.
HOLD: no unvalidated result can enter memory.
Only validated capability can alter routing, confidence, templates, or next-mission recommendations.
Every run emits a mission dossier. The standalone generates these artifacts client-side from the actual candidate output.
Confirms unrelated objectives do not produce the same jobs.