Frequently asked questions.

Runcible makes institutions operable and qualifies consequential claims before reliance. Runcible OS holds purpose, people, work, authority, consequence, and memory in one world. Runcible AI curates and tests. People retain disposition. Governed work closes with an auditable Decidability Record — a truth engine for institutional action, not another chatbot.

No. We do not train foundation models and do not compete with the labs. Models generate candidate language; Runcible tests whether that language can become institutional action. The runtime is model-neutral — commercial, private, or local.

A closed, durable artifact produced when governed work finishes: what was claimed, what evidence was used and missing, which rules and authority applied, what passed and failed, whether action is warrantable, and who is accountable. A chat transcript shows what was said. A Decidability Record shows what an institution may do.

Runcible does not return a binary accept/reject. An action state says what the institution may do next: Certified, Survives, Repairable, Escalated, Blocked, Fails, Non-warrantable, or Undecidable. Each one is read from a result state — what the test itself returned — and there are only three of those: the claim holds, it fails with an explained error, or it cannot close without discretion. Eight readings of three outcomes, because the test establishes what is true of the claim while the institution supplies what is missing: authority, evidence, scope, or liability. How the test works, and which result state each action state comes from.

Guardrails filter what a model may say. Runcible tests what an institution may do. Constraint separation — truth, reciprocity, possibility, liability — not censorship of tone.

No — it protects them. AI participants get governed roles with explicit scope and escalation. Where judgment is required, Runcible escalates to the human who holds the pen.

Start in shadow mode — review real work without changing decisions — then advisory, then human-approval. Scope expands only where records prove value. Private tenants, redacted data, and local models are supported where required.

Institutions whose work product is liability-bearing language: insurers, lenders, healthcare administrators, compliance and legal teams, government, and defense. If a wrong sentence is a lawsuit, finding, or headline — you are in scope.

The stack is long-horizon: a science of decidability developed through the Natural Law Institute, compiled into RDL, run by the governance runtime, and delivered through Runcible OS and Runcible AI. Models are the newest — and most replaceable — layer.

From Edward Lear’s “runcible spoon” — a word that sounds meaningful but has no testable definition. That is the failure mode Runcible exists to end: language that feels true but cannot be adjudicated.