Gameplay loop — from retained knowledge to trained agency [Flow · product direction]

# Gameplay loop — from retained knowledge to trained agency `[Flow · product direction]`

Meridians does not yet have one complete gameplay loop. It has a strong initial knowledge payoff, a
machine maintenance loop, a shipped Scenario turn, an incubating Episode form, and a detached learning
surface. This diagram keeps those truths separate, then defines the smallest loop worth proving next.
The shipped Scenario state machine remains owned by [Scenario loop](scenario-loop.md); World progression
remains directional under the [World contract](../../specs/world/README.md).

The analysis uses Joss Querné's four loop views as an external design lens: **play** exposes condition,
input, and output; **activities** shows the repeated cadence; **compulsion** tests motivation and
progression; **game** communicates the high-level experience. Across them, objective, challenge, and
reward must all be legible. [Source note](../../sources/external/2026-08-29-types-of-gameplay-loops.md).

## Where the loops are now

```mermaid
flowchart LR
  SOURCE["Premise · corpus · current sources"] --> EXTRACT["Extraction<br/>strong first-session comprehension"]
  EXTRACT --> DOMAIN["Maintained Domain<br/>remembered evidence + structure"]

  subgraph MACHINE["Recurring machine activity — shipped"]
    PROGRAM["Program<br/>Research → Opinion → Merge → Position"]
    STORIES["Stories + Positions<br/>material movement reaches the human"]
    PROGRAM --> STORIES -->|"next cadence"| PROGRAM
  end

  DOMAIN --> PROGRAM
  PROGRAM --> DOMAIN

  subgraph PLAY["Player-facing play — shipped core, incomplete return loop"]
    BRIEF["Brief<br/>condition + situated information"]
    PREPARE["Prepare<br/>interpret + gather Priors"]
    COMMIT["Commit<br/>express Will"]
    CONSEQUENCE["Consequence<br/>system response + preserved branch"]
    BRIEF --> PREPARE --> COMMIT --> CONSEQUENCE
  end

  DOMAIN --> BRIEF
  EPISODE["Episode / Stageplay<br/>immersion + strategic puzzle<br/>active incubation"] -.-> BRIEF
  CHAT["Chat inquiry<br/>repeatable utility, fragile consequence"] -.-> DOMAIN
  LEARNING["Learning<br/>active recall + spaced practice<br/>not joined to play"] -.-> DOMAIN

  CONSEQUENCE -. "missing hinge" .-> DEBRIEF["Debrief<br/>intention ↔ result ↔ missed signal"]
  DEBRIEF -. "not yet adaptive" .-> NEXT["Next challenge<br/>difficulty · perspective · tools · responsibility"]
  NEXT -. "return" .-> BRIEF

  classDef shipped fill:#17362f,stroke:#5fa98e,color:#eefaf6
  classDef incubating fill:#3b3020,stroke:#b59255,color:#fff8e8
  classDef gap fill:#3a2027,stroke:#bb6f7f,color:#fff0f3,stroke-dasharray:5 4
  class EXTRACT,DOMAIN,PROGRAM,STORIES,BRIEF,PREPARE,COMMIT,CONSEQUENCE,LEARNING shipped
  class EPISODE incubating
  class CHAT,DEBRIEF,NEXT gap
```

The Program loop is real, but mostly machine-side. It improves the maintained model, not demonstrably the
participant. Scenario closes one consequential turn, but the result does not yet reliably become a
debrief, an adapted challenge, or evidence of growing ability. Extraction, chat, Episode, and Learning
therefore orbit the loop rather than compounding through it.

## The next coherent game loop

The player-facing vision is: **enter a world that remembers; understand what matters from one situated
perspective; make a move that changes the world; learn from the difference between intention and result;
return able to carry greater consequence.**

```mermaid
flowchart TB
  OBJECTIVE["Contextual objective<br/>What must this seat protect, discover, change, or decide?"]

  subgraph SESSION["Activity / session loop — the next product contract"]
    RECALL["1 · Recall<br/>restore relevant history"]
    ORIENT["2 · Orient<br/>find the decision window"]
    INHABIT["3 · Inhabit<br/>role + partial information + strategic puzzle"]
    COMMIT["4 · Commit<br/>order · plan · trade · disclosure · refusal"]
    CONSEQUENCE["5 · Consequence<br/>rules + other agency resolve the move"]
    DEBRIEF["6 · Debrief<br/>intention · expectation · result · surprise"]
    ADAPT["7 · Adapt<br/>change the next challenge"]
    RECALL --> ORIENT --> INHABIT --> COMMIT --> CONSEQUENCE --> DEBRIEF --> ADAPT --> RECALL
  end

  OBJECTIVE --> RECALL

  subgraph PLAY["Inner play loop — prove moment to moment"]
    CONDITION["Condition<br/>world state + seat knowledge + available tools"]
    INPUT["Input<br/>situated Will through a permitted action"]
    OUTPUT["Output<br/>legible governed consequence"]
    CONDITION --> INPUT --> OUTPUT --> CONDITION
  end

  INHABIT --> CONDITION
  OUTPUT --> CONSEQUENCE

  MEMORY["Memory Palace + Program<br/>recall · signal selection · maintained Domain"] --> RECALL
  EPISODE["Episode<br/>embodiment · explanation · puzzle"] --> INHABIT
  SCENARIO["Scenario<br/>commitment · resolution"] --> COMMIT
  SCENARIO --> CONSEQUENCE

  CONSEQUENCE --> WORLD_RECORD[("World record<br/>what changed, who knew, which rule applied")]
  DEBRIEF --> ABILITY_RECORD[("Practice record<br/>what the participant learned to notice, judge, or do")]
  WORLD_RECORD --> MEMORY

  ABILITY_RECORD --> CAPABILITY["Capability grows<br/>comprehension · calibration · causal judgment<br/>coordination · perspective · tool use"]
  CAPABILITY --> RESPONSIBILITY["Earned progression<br/>harder problems · broader tools · new perspectives<br/>greater permission or responsibility"]
  RESPONSIBILITY --> OBJECTIVE

  SOCIAL["Other human or agent seats<br/>private knowledge + conflicting aims"] --> CONDITION
  SOCIAL --> CONSEQUENCE
```

This is a training loop, not a content treadmill. Its reward is twofold: an intrinsically useful outcome
in the Domain and increased agency in the next situation. Roles, tools, access, relationships, resources,
and responsibility can embody progression; generic XP does not prove that the participant became more
capable. The loop can work in a bounded Scenario before a persistent World exists.

## The proof ladder

```mermaid
flowchart LR
  P0["0 · Retain<br/>Extraction creates a re-enterable Domain<br/>CURRENT"]
  P1["1 · Close one turn<br/>Brief → Prepare → Commit → Consequence<br/>CURRENT"]
  P2["2 · Learn from consequence<br/>automatic evidence-backed debrief<br/>NEXT BUILD"]
  P3["3 · Adapt the return<br/>replay or next episode targets the revealed gap<br/>NEXT PROOF"]
  P4["4 · Progress through agency<br/>roles, tools, permissions, perspective, responsibility<br/>DIRECTION"]
  P5["5 · Sustain social worlds<br/>group memory, institutions, economies, long arcs<br/>CONDITIONAL"]

  P0 --> P1 --> P2 --> P3 --> P4 --> P5
```

The next stage is deliberately small: close **Consequence → Debrief → Adapt → next Brief** for one
repeatable Scenario or Episode. Test whether a participant can identify a missed signal, revise a causal
model, and perform better on a related challenge. Only then should Meridians invest heavily in persistent
RPG economies or MMO-scale retention systems.

## Design tests

| Loop view | Meridians question | Evidence of success |
|---|---|---|
| Play | Does a situated input produce a prompt, legible, governed response? | The player can explain what their move changed and why. |
| Activities | Do the seven activities form one comprehensible return cadence? | The player reaches a next challenge without navigating disconnected products. |
| Compulsion / progression | Does the reward move the player toward a meaningful desire? | A useful outcome and a specific gain in ability, access, or responsibility motivate return. |
| Game vision | Do all surfaces serve the same fantasy? | The experience consistently feels like practising agency inside a world that remembers. |

Every challenge needs both a **contextual goal** for the participant and a **systemic goal** the engine can
resolve. “Prevent the coalition from fracturing” carries the dramatic intent; “secure two commitments
before the vote under the seat's information and resource constraints” makes the challenge playable and
testable.

Raw Markdown source