Runcible: The Other Half of AI
In simple terms, Runcible is a truth engine.
LLMs generate hypotheses. Runcible subjects them to attempted falsification—examining what is claimed, what it depends on, and whether it withstands scrutiny. It constrains the AI to an explicit method of examination, limiting its freedom to substitute interpretation for demonstration.
Runcible is a general-purpose adjudication engine. Its method applies to expressions, claims, assertions, arguments, plans, policies, and entire bodies of knowledge. It exposes ignorance, error, bias, and deceit; explains what can be demonstrated and defended; identifies what remains unresolved; and recommends means of correction, reconciliation, or settlement.
Runcible began as an implementation of the science of decidability applied to constitutional law. We used it in our work advising the high courts through amicus briefs. That demanding application established its original purpose; the method extends across fields wherever information must be examined and decisions justified.
That is why Runcible is “The Other Half of AI”: generation supplies possibilities; adjudication establishes which can withstand examination and on what grounds.
Oversing: The institutional operating system
Think of it as the organization’s central nervous system—connecting its memory, rules, people, and work.
Oversing brings strategy, staffing, customers, projects, and financial commitments into a shared institutional context. It puts Runcible’s findings to work by connecting decisions to the responsible people, the resources required, and the rules governing action.
Its architecture separates what AI proposes from what the institution authorizes and executes. The record preserves what happened so the organization can learn from the results.
The work is already there. So is the economic opportunity.
Organizations spend heavily establishing facts, evaluating proposals, resolving disagreements, coordinating commitments, and correcting mistakes.
Generative AI makes producing candidate answers inexpensive. The institution still carries the responsibility for determining what holds, what can be done, whose interests are affected, and who may act.
Runcible addresses that continuing burden: the work required to turn information into a defensible decision and carry it through an organization.
Across law, insurance, healthcare, finance, enterprise management, and government, the opportunity reaches from better judgment to more effective institutional cooperation.
Explore the institutional AI opportunity →A separation of powers for enterprise AI.
Two products connect intelligence to institutional authority. The architecture assigns generation, decision, and execution to distinct responsibilities.
Generate and examine.
The foundation model and Runcible construct and test candidate claims, preserving findings, conditions, and repair options.
Decide and authorize.
Oversing supplies the institution’s procedures, responsible authorities, permissions, and delegation boundaries.
Execute within the mandate.
Oversing issues an authorized task to a separate executing model or tool. The generating model cannot turn its recommendation into permission.
Authority remains with the institution as AI capability expands.
See how the architecture works →Examine the claim. Find a way forward.
A truth engine needs to show its work. Runcible tests what a claim depends on, explains where it holds or fails, and develops alternatives when it cannot be supported.
Runcible reconstructs what a claim requires—the actors, operations, evidence, resources, obligations, and constraints—and attempts to defeat it through contradiction, impossibility, missing prerequisites, or unsupported relations.
The examination addresses four distinct burdens:
Possibility
Whether the proposed account or action can be constructed under its conditions.
Ethics and morality
Whose interests are affected, what obligations apply, and where costs or risks are transferred.
Truth and testifiability
What the evidence supports, what survives challenge, and what remains unresolved.
Warrantability and liability
What reliance the organization can undertake, who is responsible, and what follows if it fails.
Factual findings remain distinguishable from the policies, interests, and preferences governing a decision.
When a proposal fails, Runcible identifies the defect and develops alternatives: additional evidence, corrected terms, compensation, compromise, or a different course of action.
Explore Runcible AI →Put the institution in charge of AI-assisted work.
A sound recommendation still needs context, responsibility, and permission before it becomes institutional action.
Oversing retains the institution’s history, decisions, commitments, and operating rules. It connects each proposal to the people responsible, the resources available, and the procedures that govern what happens next.
A recommendation enters that institutional context. The decision belongs to a named authority or an explicit delegation policy; the resulting work carries its permissions and responsibilities with it. A model’s recommendation cannot authorize itself.
The shared record connects decisions to their consequences, so people can review what happened, reconcile competing interests, and improve the next decision. The institution retains what it learns.
Explore Oversing →Catch the defective proposal before it becomes institutional action.
An executive proposes pooling sales commissions already earned to improve teamwork. Producing a policy is easy. Establishing whether the organization can properly adopt it requires examining existing agreements, affected interests, authority, and the claimed benefit.
In the recorded case, Runcible distinguishes earned compensation from future terms and voluntary contributions, then develops alternatives under the supplied agreements. The decision-maker receives a specific defect to address and options for preserving the legitimate objective.
The accompanying institutional walkthrough shows how those findings would connect to approval, responsibilities, implementation, and review in Oversing.
Follow the compensation case →Keep the examination. Build on what it teaches.
Runcible retains the grounds of an examination: the question, evidence, assumptions, attempted falsifications, surviving distinctions, unresolved conditions, and proposed repairs.
Reviewers can inspect why a proposal failed, what information is missing, and what would permit a different conclusion.
Oversing supplies the continuing institutional context. Connecting examinations with decisions and observed consequences creates the basis for better protocols, reusable implementation knowledge, and controlled model improvement.
People learn the method through their work. The organization retains that learning beyond an individual conversation or employee.
Built for enterprise delivery.
For Systems Integrators
Bring semantic examination and institutional authority into your clients’ consequential workflows.
Start with a bounded matter, a named executive owner, and measurable operating requirements. Establish the implementation pattern, measure its results, and expand into adjacent work.
Partners contribute domain expertise, integration, configuration, and organizational change—and earn the associated implementation and service revenue. Runcible supplies the software and method, with partner tooling, deployment playbooks, training, and support forming the commercialization program.
Explore implementation partnerships →For enterprise and government leaders
Evaluate the platform against a recurring decision, dispute, or coordination problem.
Inspect its findings, authority boundaries, and fit with your existing procedures. Identify the responsible participants, the operating baseline, and the result your organization needs to achieve.
Discuss an enterprise evaluation →Working software. A substantial foundation.
Runcible brings together an operating AI runtime, implemented protocols, retained examination records, and a deployed multi-user Oversing application.
The method draws on the Natural Law Institute’s research into language as measurement, constructive falsification, reciprocity, and institutional cooperation.
We are packaging that foundation for repeatable enterprise deployment through Systems Integrators. Technical evaluators can inspect the architecture, current implementation, and engineering milestones.
Start with one consequential problem.
Bring a proposal, agreement, disputed claim, or recurring operational decision.
We will walk through the examination, inspect the resulting record, and assess how the work connects to your institution’s people, procedures, and authority.