A language model can interpret a situation, propose an action or generate dialogue. That does not make the described event true.
If an Actor says “I’ll go,” CPS does not treat that sentence as evidence that they moved. A decision, a commitment, scheduled execution, movement, an encounter and its outcome are distinct states.
CPS keeps those states separate so later behavior can depend on what actually happened rather than what a model merely said happened.
CPS separates world truth from what individual Actors know, believe, suspect or remember. Information remains Actor-local and retains its source, confidence and provenance, allowing direct observation, testimony and uncertain belief to remain meaningfully different.
Possibility is constrained by the current world: time, location, activity, schedule, existing commitments, relationships, accessible information and other relevant context. CPS determines which opportunities are legitimate rather than allowing a model to freely invent what happened next.
An opportunity does not automatically become an action. An Actor can respond, refuse, defer, continue an existing commitment or do nothing. When a decision requires future execution, CPS can retain it as durable state rather than relying on conversational prose to remember it.
CPS keeps intention separate from execution. Movement must occur in the world. Encounters must actually take place. Actions can succeed, fail or remain incomplete. Only legitimate outcomes become evidence that later memory, relationships and behavior can depend on.
The flow shows how world state can become consequence without implying that every observation creates an intention or that every decision becomes an action.
Something exists or changes in the canonical world.
What can this Actor legitimately perceive, receive or retrieve?
What, if anything, becomes significant to this Actor?
What does the Actor make of it, and is any response warranted?
The context may be irrelevant, or the Actor may continue what they were already doing.
The matter can remain pending and return through later context without becoming immediate action.
A decision that requires future execution becomes durable state rather than conversational implication.
The execution path continues only when a legitimate commitment exists
Timing, current activity and existing obligations determine whether and when work is attempted.
The action must occur in the world. Movement changes location; encounters require legitimate co-presence.
The result becomes evidence only when the required causal steps actually occurred.
What Actors know, remember, believe and how they relate may change as a consequence.
The consequence becomes part of what may legitimately matter in a future situation.
Later context returns to the world loop. No action, deferral and failure remain valid parts of that history.
A persistent artificial life is not one continuous decision loop. Most things do not become actions. CPS preserves the intermediate states so that consequences occur only when the required causal steps actually happened.
Something becomes available to an Actor through the world, communication, memory or another legitimate source. Availability does not guarantee that the Actor notices it or acts on it.
Accessible context may become relevant to the Actor’s current situation. The result can be an action, a deferred intention, continued existing behavior or no response at all.
When something must happen later, CPS can preserve the obligation beyond the conversation that created it. Scheduling and world execution determine whether the intended action is actually attempted.
Completed actions create canonical evidence. Outcomes may change what Actors know, remember, believe, how relationships evolve, where Actors are, what commitments remain, and what later situations become possible.
Knowledge in CPS is not automatically global.
When one Actor learns something from another, the recipient receives their own epistemic record rather than inheriting the source Actor’s experience as objective truth. CPS can retain who supplied the information, whether it was directly observed or reported, and how it relates to earlier evidence.
Whether that information becomes relevant again depends on later circumstances. It may influence another exchange, remain private or irrelevant, be contradicted by new evidence, or simply never matter again.
Propagation is therefore a consequence of legitimate social circumstances, not the default destination of every memory.
CPS Studio makes causal claims inspectable.
An operator can trace what an Actor had access to, what became relevant, what was interpreted, what was decided, which obligations existed, what was scheduled, what actually executed, and what canonical state changed afterward.
Speech, intention, schedule and outcome remain distinguishable.
This makes it possible to inspect not only what the system produced, but why later behavior is allowed to depend on it.
A model may interpret context, propose a response or generate dialogue. CPS remains responsible for the structured state surrounding that output: world facts, Actor-local knowledge, decisions, commitments, timing, movement, encounters and their resulting evidence.
A language model can describe an event.
CPS determines whether that event actually became part of the world.