Technical Whitepaper / CPS Core v2

Context
Propagation
System

Causal continuity infrastructure for artificial worlds.

Wojciech Hupert
Version 1.0 / August 2026
12 pages / PDF

Abstract

A world should not have to invent its past every time a model runs.

The Context Propagation System is an operational framework for persistent, inspectable artificial worlds in which information has social and causal consequences. It connects shared World state, actor-local Memory and interpretation, social interaction, action arbitration, external model providers, and human-facing observability through one durable causal loop.

CPS treats language models as bounded realization and interpretation providers, not as authorities over world truth. Objective location belongs to World. Subjective memory and meaning belong to Anima. Interaction legitimacy belongs to Forum. Work selection belongs to Scheduler. Observatory explains the result without mutating it.

Events become evidence. Evidence becomes actor-local meaning. Meaning can become action. Accepted consequences remain available to shape what happens next.

Canonical Causal Loop

One traceable path from event to future consequence.

01StimulusWorld event, conversation, Player input, or accepted consequence.
02Actor-local meaningAccess, attention, perception, interpretation, and reassessment.
03Disposition or workA bounded reason becomes a runnable candidate or durable Task.
04ArbitrationScheduler selects, defers, blocks, or rejects competing work.
05Domain executionForum governs interaction; World governs shared physical truth.
06RealizationA model or device supplies bounded output or execution evidence.
07Canonical consequenceThe correct authority validates and persists what actually changed.
08Future causeThe accepted result remains available to influence later behavior.

Inside the paper

Architecture, operation, evidence, and application.

01-03

The continuity problem

Why fluent dialogue and long context do not by themselves create a persistent world.

04-05

Six pillars and one action spine

Core, World, Anima, Forum, Soma, Observatory, and shared Scheduler arbitration.

06-09

Memory, Reflection, social life, and artifacts

How retained experience can alter later behavior and produce source-grounded work.

10-12

Providers, Studio, and reliability

Non-authoritative models, progressive disclosure, persistence, replay, and recovery.

13-14

Lelit, embodiment, and evaluation

The first operational world, the Unreal boundary, and causal acceptance evidence.

15-18

Applications, context, and limits

Where CPS can matter, how it relates to adjacent work, and what it does not claim.

System position

Providers are expressive. CPS remains authoritative.

A provider can realize selected dialogue, extract bounded Memory content, propose a source-grounded Reflection, or generate an attributed artifact. It cannot decide that a World event occurred, move an actor by prose, complete a Task, create a relationship consequence, or bypass admission and arbitration.

Operational world

Lelit Distrikt makes continuity visible.

The bounded Berlin district gives CPS authored residents, routines, relationships, locations, conversations, actor-created work, and enough simulated time for consequences to return. CPS Studio lets a person participate, inspect the world, and follow direct causal lineage.

Scope

A working bounded system, not a universal society simulator.

Bounded is a design property. CPS operates through authored identities, validated data, explicit policies, typed transitions, and inspectable evidence. Silence, no-change, uncertainty, blocked action, and provider failure are truthful results. The system does not claim consciousness, unrestricted autonomy, or human-equivalent social understanding.

Read the full whitepaper

Published 24 August 2026 / Version 1.0 / Context Propagation System