HEAVEN
A persistent world for cognitive digital twins.
HEAVEN is a deterministic simulation platform where agents preserve memory, evolve through traceable decisions, and inhabit replayable worlds.
- 01 First world
- EDEN bounded demo
- 02 Continuity
- Memory across time
- 03 Behavior
- Traceable and replayable
HEAVEN
The platform layer: persistent cognitive agents, replayable worlds, and observable continuity in one product experience.
PlatformThe World
The visible environment where persistent agents act, accumulate history, and become inspectable over time.
Product ExperienceThe Engine
The technical engine behind HEAVEN: deterministic execution, event-sourced state, decision traces, and replayable artifacts.
RuntimeMemory
The continuity layer that transforms events and interactions into durable, behaviorally relevant memory.
ContinuityUse Cases
Applications across persistent cognitive agents, simulation, digital twin behavior, research tools, and careful memory continuity experiences.
ApplicationsResearch
The theoretical and architectural foundation behind HEAVEN: memory, decision, movement, logic, and metacognitive depth.
TheoryHEAVEN
HEAVEN is the product layer: a persistent cognitive digital twin platform where agents keep continuity, generate observable behavior, and remain grounded in deterministic traces.
- Persistent agentsDigital twins with durable state, identity traces, and behavior that can be inspected over time.
- Replayable worldsWorld experiences designed around continuity, causality, and visible change.
- Traceable behaviorEvery meaningful decision is connected back to state, memory, and runtime context.
The World
EDEN is the first bounded demo world: a visible environment where persistent agents move through time, accumulate history, and make continuity legible.
- t+001 Agent enters EDEN with a bounded identity trace.
- t+042 Memory provenance links action, place, and outcome.
- replay The world state can be inspected as a continuity record.
- EDENA contained world for demonstrating persistent environments and agent continuity.
- Visible stateMovement, memory, and outcomes are surfaced as part of the product experience.
- Long-running contextThe world remembers enough to make change feel cumulative rather than episodic.
The Engine
heaven-engine is the technical runtime underneath the platform: deterministic ticks, event-sourced state, structured artifacts, and replayable behavior.
- Deterministic runtimeWorld state advances through controlled ticks instead of opaque session drift.
- Event-sourced stateArtifacts preserve what happened, when it happened, and which systems shaped the result.
- ReplayabilityBehavior can be inspected through traces rather than treated as a black box.
Memory
Memory is the continuity layer that turns events, interactions, and outcomes into durable context for persistent cognitive agents.
- ProvenanceMemories stay connected to their source events and world conditions.
- Continuity over timeAgents can carry context forward without implying metaphysical persistence.
- Behavioral relevanceMemory becomes useful when it changes what an agent can notice, choose, or revisit.
Use Cases
HEAVEN is built for careful applications where long-term agent continuity, simulation, and digital twin behavior need to be observable and replayable.
- Research simulationStudy persistent cognitive agents inside bounded worlds with inspectable state.
- Digital twin behaviorModel identity traces, preferences, routines, and decision patterns with explicit limits.
- Legacy and memory toolsExplore continuity experiences based on records, interaction history, and transparent simulation boundaries.
Research
The internal theoretical architecture connects memory, movement, decision, logic, and bounded metacognitive depth across the HEAVEN platform.
- CodexThe organizing layer for research notes, architecture, and implementation intent.
- AOMDAction-Oriented Metacognitive Depth: a theory of deliberative closure, reflective depth, and bounded action-oriented reasoning.
- MAV1 / MoAV1Memory and movement architectures for persistent agent behavior.
- LPAV1Logical Protocol Architecture for constraints, validation, and inspectable agent action.