Skip to content

Starter Pack UX Schematic — First 20 Games

Status: Proposal for review by CC. Not committed. Date: 2026-06-01 Authors: CoachJ + Claude Supersedes (in part): ADR-073's level-1-through-4 structure for games 1–20. The core insight from ADR-073 — don't show coherence % early; use a stars-style game-count as the visible progression instead — is preserved. The level partitioning is replaced by world-based progression.

At a glance

What it proposes: a "hub + visible locked worlds" UX for the first ~20 games, with onboarding-picks-interests personalizing which worlds open first. Built on a Mechanics→Dynamics→Aesthetics (MDA) chain so the design choices are derivable from goals, not just intuition.

  • Aesthetic goal: Revelation + Self-discovery. The player should feel "this AI is starting to see me" and "I'm finding out what I actually value."
  • Core mechanic: Hub with 7 surface "worlds" (career, health, relational, financial-personal, civic, time-energy, creative-strategic). Onboarding asks the player to pick 3–4 worlds they face tough decisions in; those open first, others appear as locked.
  • Game 1: always the same curated tutorial scenario ($20 Loan to Sam), surface-agnostic, lives in a "lobby" before worlds open.
  • Progression mechanism: count-based for inside-a-world unlocks ("play 3 in a world to deepen it"), event-based for first-MCP-reflection moment ("first time Claude calls your profile, a new world unlocks as celebration"). ADR-073's per-game stars counter replaces coherence % until ~game 20.
  • Why this shape: answers the "what's meaningful for MCP" question structurally (player picks → those are the meaningful surfaces) and provides cross-domain demonstration through gameplay rather than scripting.

Why a methodology and not just intuition

CoachJ's framing at session close: "How are we deciding what's the schematic we're gonna use to make the decisions of how to do this and how to reveal it and what scenarios to choose and what surface areas to include? I'm approaching this intuitively again. And I wonder if there's a more structured way that this can be designed and approached with."

Right impulse. Game design has standard frameworks for exactly this. The one used here is MDA — Mechanics → Dynamics → Aesthetics, from Hunicke / LeBlanc / Zubek. Design backwards from how you want the player to feel, not forwards from "what's a cool feature."

The methodology in five steps:

  1. State the aesthetic goal — what should the player feel
  2. Identify dynamics that produce that aesthetic — what kinds of in-game behavior generate the feeling
  3. Pick mechanics that produce those dynamics — what concrete systems / surfaces / rules
  4. Decide progression structure — how access expands over time
  5. Reconcile with technical + product constraints — what's actually buildable / measurable

Walked through for Pulse first-20-games below.

Step 1 — Aesthetic goal

Pulse's marketing thesis points at two aesthetics:

  • Revelation — "this AI is starting to see me"
  • Self-discovery — "I'm finding out what I actually value"

Explicitly not the goal:

  • Mastery — would cut against the judgment-elicitation thesis. There are no "right answers" to get better at. Framing it as Mastery makes the player look for the answer the AI wants instead of revealing their own judgment.
  • Competition — same problem. Pulse is single-player against your own future judgment, not against other players' scores.

Step 2 — Dynamics that produce those aesthetics

For Revelation to emerge: - Insights at reveal time that name patterns the player hadn't articulated - Claude MCP reflecting back judgment patterns the player can recognize - AI predictions that get measurably more accurate as the profile builds

For Self-discovery to emerge: - Variety of scenarios across surfaces, so the player sees themselves in different decision contexts - Agency in which surface to explore next — choice itself is signal - A visible map of "where I've been" so the player can see the shape of their own engagement

Step 3 — Mechanics that produce those dynamics

Dynamic needed Mechanic
Variety + agency in surface choice Hub with worlds; player picks within available set
Visible "where I've been" Hub shows played/unplayed worlds and counts
Insights that name unarticulated patterns Reveal-time insight engine (existing)
Claude MCP reflecting back MCP integration (existing)
AI predictions getting more accurate Pulse's existing prediction pipeline (existing)
Cross-domain demonstration Multiple worlds available, structurally encouraging multi-surface play

Step 4 — Progression structure

Three patterns from game design:

Pattern Example When it fits Pulse
Count-based ("play 3 to unlock") Mario 64 stars Inside-a-world progression. Predictable, controllable.
Quality-based ("complete at level X to unlock") Most JRPGs Doesn't fit Pulse — there's no mastery tier.
Event-based ("when X happens, Y unlocks") Stardew Valley seasonal events First MCP-reflection moment. The event IS the reward.

Recommendation: Hybrid — count-based for inside-a-world depth, event-based for the first-MCP-reflection celebration.

Step 5 — Reconciled schematic

flowchart TB
    O[Onboarding] --> O1[Connect Claude MCP<br/>skippable, nudged]
    O --> O2[Pick 3-4 of 7 surfaces<br/>'where do you face tough decisions?']
    O --> G1[Game 1: $20 Loan to Sam<br/>always curated, surface-agnostic]
    G1 -->|triggers| MCP[MCP-reflection event<br/>1st locked world unlocks]
    G1 --> HUB[Hub]
    O2 --> HUB
    MCP --> HUB
    HUB --> W1[3-4 worlds unlocked from onboarding picks]
    HUB --> W2[3-4 worlds visible but locked]
    W1 --> Play[Play scenarios in any unlocked world<br/>3-5 available per world, filtered from corpus]
    Play -->|play 3 in any world| Unlock[Unlock 1 locked world]
    Unlock --> HUB
    Play -.->|per-game stars counter| Stars[Visible progression: # of scenarios played<br/>NOT coherence % per ADR-073]

Concrete rules:

  • Onboarding asks the player to pick 3–4 of 7 surface tiles ("Which areas do you face tough decisions in most often?")
  • Game 1 is fixed: the $20 Loan to Sam curated tutorial scenario. Lives in a "lobby" before the hub opens.
  • Game 1 completion triggers MCP-reflection event. One additional locked world unlocks ("you've shown your AI what you value — explore something new").
  • Hub displays all 7 worlds. Unlocked ones show 3–5 available scenarios (filtered from cube-tagged corpus for that surface, varied across difficulty). Locked ones show as aspirational tiles with "unlock after 3 plays in any world."
  • Inside a world, the player picks any of the available scenarios. After completing them, new scenarios refill the slot (corpus is large enough — and growing — to support replays without immediate repetition).
  • Visible progression metric is the per-game stars counter (preserved from ADR-073's core insight). Coherence % is hidden until ~game 20.
  • After ~game 20, all worlds are unlocked, the hub stabilizes, and coherence % becomes visible. This is the transition from starter pack to free play.

Why this answers the harder questions

"What's meaningful for MCP?"

The player's onboarding picks ARE the meaningful surfaces. Pulse stops guessing. When their Claude MCP gets a call about a career decision, the profile has career-surface judgment data because the player chose to explore that world.

"What's the right curated footprint?"

Just Game 1. Everything else can come from the (expanded) corpus filtered by surface + cube vector. The earlier "curate Games 2–5" question dissolves because Games 2+ are whatever world the player picked first, served from corpus.

"How many surface areas?"

7 as the master taxonomy (matching the cube + generation prompt). The "could be 6, could be 10, could be 50" question doesn't need to be resolved upfront — the player only ever sees 7 tiles, but the corpus behind each can be as deep as we author.

"When does the player feel the MCP value?"

The first MCP-reflection event after Game 1 is the structural moment. It unlocks a new world as visible celebration. The Claude reflection on the player's $20-loan choice becomes the proof point.

"How does this not feel like a test?"

Agency. The player is choosing worlds, choosing scenarios within worlds, choosing when to engage. Forced sequence reads as a test; chosen exploration reads as a game.

What's still genuinely a designer's call (CC's territory)

These are not derivable from the methodology — they're real preference + craft choices:

  • 3 picks vs 4 picks at onboarding. 3 = more focused; 4 = more variety from the start.
  • Count threshold for unlock. "Play 3 to unlock" vs "5 to unlock" — affects pace.
  • Locked-world visual treatment. Mario 64 style (visible locked doors, obviously aspirational) vs Stardew Valley style (appear on unlock, surprise + delight).
  • MCP-reflection event presentation. Celebratory animation vs subtle notification.
  • Per-game stars counter visual. Literal stars? A filling bar? Just a number?
  • Order of onboarding steps. MCP-connect before world-picks? After? Skippable in different orders?
  • Whether the player can change their world picks later. "Re-rank your interests" as a settings option vs locked-in at onboarding.

Open questions to settle with CC

  1. Master surface taxonomy — confirm 7 surfaces? Current candidates: career, health, relational, financial-personal, civic, time-energy, creative-strategic. Could be different.
  2. MCP-reflection event mechanics. Does the unlock happen automatically when Claude makes its first call referencing the profile? Or does the player trigger it manually ("I tried it in Claude — show me what unlocked")?
  3. Re-pickability of onboarding choices. Should players be able to change their initial 3–4 picks after game N? If yes, what triggers it?
  4. First-MCP-reflection chicken-and-egg risk. If a player skips MCP-connect at onboarding, when does the first MCP-reflection event fire? Possible fallback: a synthetic "preview" reflection generated locally so the player sees what the unlock moment looks like even before connecting.
  5. What happens at game 20+ when all worlds are unlocked. Does the hub change? Does coherence % gradually reveal? How do reactive arcs / campaigns slot in?

What changed vs ADR-073

ADR-073 This proposal
Levels 1–4 across games 1–20 World-based progression; no levels
Linear progression through curated scenarios Hub with worlds; player chooses within unlocked set
Curate first 5–6 scenarios Curate only Game 1; Games 2+ from filtered corpus
Per-game stars counter (preserved) Per-game stars counter (preserved)
Coherence % hidden until ~game 20 (preserved) Coherence % hidden until ~game 20 (preserved)
MCP-connect after Game 1 (preserved) MCP-connect at onboarding (slight shift earlier — could revert if CC prefers)
Personalization of starter pack: not addressed Onboarding picks personalize world unlock order

The supersession is partial. Core ADR-073 insight (per-game counter early, coherence later) is the foundation. Level partitioning is replaced because world progression doesn't naturally divide into 4 levels of 5 games each.

If this proposal is accepted, ADR-073 should be amended (or superseded with a new ADR) to reflect the world-based structure.

Implementation order (suggested)

If this proposal is accepted:

  1. Author the master surface taxonomy doc — define the 7 surfaces with one-line descriptions for the onboarding tile copy
  2. Generate corpus expansion in parallel — already in flight via the generation prompt; the corpus needs to populate the 7 surfaces with cube-vector spread
  3. Build the hub UI — CC's domain; world tiles + scenario picker within a world
  4. Build the onboarding flow — surface-picker + skippable MCP-connect + Game 1 entry
  5. Build the unlock + MCP-reflection event — count tracking + event firing
  6. Author the Game 1 scenario ($20 Loan to Sam) — already drafted in v3-redesign-draft.md; needs Bias 6/7 verification + ship
  7. Wire up the per-game stars counter — replace coherence % display for game < 20

Most of this is CC + James implementation work. The cube + corpus work has been the bottleneck; that's now unblocked.

Reference

  • ADR-073 — original starter pack design (partially superseded)
  • ADR-074 — Pulse as multi-domain judgment layer (the strategic basis for surfaces)
  • docs/research/foundations/scenario-dimension-cube.md — the cube model
  • docs/research/scenario-generation-prompt-2026-06-01.md — generation prompt for corpus expansion
  • docs/starter-pack/v3-redesign-draft.md — Game 1 draft ($20 Loan to Sam)
  • docs/research/audits/corpus-cleanup-status-2026-06-01.md — corpus state after cleanup pass
  • MDA framework — Hunicke, LeBlanc, Zubek 2004