The Session: Time Windows in Persistent World Systems
In most interactive systems, time is a property of the session.
The world begins when the session begins. Content is initialized, state is loaded or generated, the interaction proceeds. When the session ends, the world — in any meaningful dynamic sense — ends with it. What persists is a saved record, a static snapshot available for restoration. The world itself does not continue.
This model encodes a specific ontological assumption: the system's live dynamics are coextensive with the session. Nothing runs between sessions. The world does not have a state independent of an active interaction.
Persistent world systems invert this relationship.
In a persistent world system, the session is a property of time. The world exists continuously, independently of whether any session is active. A session is a bounded temporal interval during which an external agent gains observation access and action capability. When the session ends, the world retains its current state. It does not pause, reset, or expire. It holds.
A Formal Definition
A Session is a bounded temporal interval with three structural properties.
Observation access. During a session, the current state of the world is made available to an external agent. The agent receives a report of the world's present configuration — its active tensions, accumulated events, character states, and causal history as encoded in the current state vector.
Action capability. During a session, the agent may introduce Wand actions into the system. These actions perturb world state according to the dynamics described elsewhere. The session is the interval during which such perturbations are possible.
Temporal boundedness. A session has a beginning and an end. These boundaries are properties of the agent's connection to the world — not properties of the world itself. The world existed before the session began. It continues after the session ends. Its state at the start of a new session is identical to its state at the close of the previous one.
The session is not a container of the world's dynamics. It is an aperture through which those dynamics are observed and perturbed.
Resettable Time vs Continuous Time
Two distinct temporal models govern interactive systems.
In resettable time, the system has a privileged initial state to which it can be returned. This initial state functions as a temporal anchor: "Round 1," "New Game," "Fresh Start." The world can be rewound. Each playthrough, each session, each invocation may begin from the same conditions. Time is indexed rather than continuous — there are positions within it that can be revisited.
Resettable time is appropriate to systems organized around mastery and iteration. The ability to restart from known conditions allows users to learn, practice, and improve. The world is instrumental — a designed environment for the agent's development.
In continuous time, no privileged initial state exists to return to. The current state is the only state. Earlier states are not preserved for restoration — they are encoded, in causal residue, in the current state's accumulated structure. You cannot return to an earlier time because the system at an earlier time no longer exists. What exists is the present state, which is the product of everything that has passed through the system.
Continuous time is the only model compatible with persistent world dynamics. A world that can be rewound is not persistent — it is a resettable simulation with persistence features. The temporal irreversibility is not an implementation detail. It is the defining structural commitment.
Reset and the Collapse of Dynamics
The dynamics of a persistent world are entirely a property of accumulated state.
At Round 3, the system contains three rounds of causal history: the events that occurred, the relationships that formed, the tensions that were activated or avoided, the weights that shifted. The narrative behavior at Round 3 emerges from this specific configuration.
At Round 30, the system contains thirty rounds. The behavior at Round 30 is a function of this configuration — qualitatively different from Round 3 not because the rules changed, but because the state encodes a fundamentally more complex causal structure.
A reset does not return the system to Round 1. It initializes a new instance with no accumulated history. The configuration that produced Round 30's behavior does not exist in the reset system — not in a dormant form, not in a latent form. It is absent. The dynamics that emerged from that configuration are absent with it.
This is the dynamics collapse argument: reset eliminates the accumulated state that the system's behavioral complexity is a function of. What survives is the World definition — the rules, the entity types, the causal structure — but not the specific Instance's history. The reset system is not the same system at an earlier time. It is a new system that shares a definition with the old one.
Resetting a persistent world does not restore it. It starts a different world that happens to share the same governing structure.
Session as Observation Window
An analogy from scientific practice clarifies the structural relationship.
A geologist extracts a core sample during a field session. The geological processes that produced the core — deposition, compression, heat, time — operated continuously across spans of time that dwarf the session itself. The session gives the geologist access to the current evidence of those processes. When the session ends, the geological processes do not pause. They continue, indifferent to the observation.
The session was an aperture, not an enclosure. The geologist observed a continuous system during a bounded interval. The system's dynamics were not contingent on the observation.
Sessions in persistent world systems have the same structure. The world's dynamics — its accumulated state, its causal history, its ongoing tensions — are not contingent on an agent's presence. The session opens a window through which the current state is readable and Wand actions are introducible. Between sessions, the state holds: not advancing (since advancement requires action), but not expiring either.
The world is not running between sessions in the sense of autonomous advancement. It is persisting — maintaining its current configuration, available for the next session to encounter exactly as the previous session left it.
Cognitive Implications
The dominant cognitive model for interactive systems treats session boundaries and world boundaries as coincident.
Under this model: the world begins when you enter it, exists while you are in it, and ends or suspends when you leave. Re-entry either restores the world to a saved state or begins a new instance. In either case, the agent's presence is constitutive — the world is defined by its relationship to an active session.
In a persistent world system, this model generates characteristic errors.
The most common: treating session end as world suspension. The agent closes the session expecting the world to hold its breath — to exist in a kind of temporal limbo until the next session opens. This is structurally inaccurate. The world is not suspended; it is persisting in its current state, which is a specific configuration with no relationship to the session's temporal status.
The second: treating session start as world re-initialization. The agent opens a new session expecting the world to present fresh conditions, as though the previous session's effects have been processed and a new cycle has begun. In a persistent world, no such cycle exists. The session opens into the exact state the previous session left.
Both errors follow from treating the session as a container of the world's dynamics rather than as a window into them.
Agency Across Time
In session-bounded systems, agency is instantaneous: what you do in a session is complete when the session closes. The effects of your actions are contained within the session's temporal scope, or archived in a fixed record.
In persistent world systems, agency is temporal: the effects of an action introduced in one session are present in subsequent sessions, modified by intervening dynamics but causally traceable. An action at Round 5 is a structural ancestor of the world's configuration at Round 40. The agent who introduced that action in an early session has influence that extends across sessions they have not yet had.
This extension of agency across time is not metaphorical. It is a structural consequence of state continuity. Accumulated state encodes causal history. Causal history includes the agent's prior actions. Those actions remain present in the current configuration — reduced in direct influence by the event decay mechanics, but not erased.
The agent who returns to a world after an extended absence is not returning to the world they left. They are returning to a world that has held the state they created — and they left evidence there. The session they are entering inherits everything the previous sessions produced.
This is a different model of temporal agency from the instantaneous model. In the instantaneous model, action is complete when the session closes. In the temporal model, action is never fully complete — it continues to propagate, at decreasing weight, through all future states.
A session does not hold the world; it meets it.
For the state accumulation structure that persists between sessions, see Instance Identity in Persistent World Systems. For the perturbation model that determines how session actions propagate, see The Wand: Influence in Persistent World Systems.
