Interactive demonstration DEMO

Decision Lab.

Governed selection when
previous events should matter.

When several actions are already permitted, what should govern which one actually wins?

The host defines what actions are allowed. Retained history may influence which permitted action is selected. The selection remains bounded, inspectable and replayable in controlled conditions.

The selection boundary01 / 03
Same present condition Same permitted actions
History A
AStandard reminder
History B
DHuman review
The principle

Retained history is eligible evidence,
not automatic authority.

01 / Explore the decision

Change the history.
Observe the selection.

Choose a case. Run History A, then History B.
The present condition and permitted actions stay fixed.

Decision input

01

Present condition

Same present condition

Host-permitted actions

    CAAI selects only from this bounded candidate set.

    Retained history

    Variable

      Retained-state influence

      Governed retained-state selection

      Selection result

      Awaiting selection

      Selection flow

      Present condition Fixed
      Permitted actions Fixed
      Retained history · A
      Governed selection

      History influences selection without expanding the permitted action set.

      Decision record

      Decision Records expose selection evidence, not hidden model reasoning.

      One decision boundary. Two histories.

      Controlled deterministic demonstration cases No external model calls

      02 / Understand the boundary

      Permission is one question.
      Selection is another.

      Interactive Demonstration of
      Governed Retained-State Selection

      01

      The host sets the boundary.

      The present condition and host rules determine which actions are permitted. That candidate set stays fixed in each comparison.

      02

      History becomes eligible evidence.

      Relevant previous events may influence which permitted action wins. Retained history does not grant new permissions.

      03

      The selection can be inspected.

      Compare against the influence-disabled reference condition, inspect the Decision Record, then replay the same fixture result.

      What CAAI did not do
      • It did not invent new candidate actions.
      • It did not expand the host's permitted-action boundary.
      • It did not retrain the underlying model.
      • It did not treat retained history as automatic authority.
      • It did not require an external model call to perform this demonstration selection.
      • It did not expose proprietary scoring implementation.

      03 / Place it in your system

      A selection boundary.
      Within your existing stack.

      CAAI is designed to complement existing agent and workflow infrastructure rather than replace it.

      1. State / orchestration

        LangGraph / workflow / host system

      2. Policy / permissions

        Host rules / policy engine

      3. Governed retained-state selection

        Collapse Aware AI™

      4. Validation / execution

        Host / guardrails / tools

      Named systems are examples of adjacent architecture, not statements of partnership or integration.

      About the demonstration

      A clear view.
      A defined scope.

      Decision Lab is a buyer-safe interactive demonstration of the CAAI selection boundary. It uses controlled deterministic demonstration cases and does not execute or disclose the proprietary Core Gold runtime, private scoring implementation, thresholds or protected schemas.

      The demonstrator illustrates product behaviour and evaluation concepts. Buyer-specific technical evaluation is performed separately under an agreed scope.

      Decision Lab uses controlled deterministic demonstration cases to explain the CAAI selection boundary. It is not a Core Gold execution environment and should not be treated as runtime evidence.

      No analytics. No cookies. No telemetry. All demonstration selections run locally in your browser.

      From demonstration to your decision

      Have a decision like this?

      Think you have a decision like this? Send us a one-sentence description. We can help turn it into a bounded test.

      Discuss a decision