Prompt Deck Spec
How do we build an instrument that gets smarter each time someone uses it?
Build Contract
| # | Feature | Function | Outcome | Job | State |
|---|---|---|---|---|---|
| 1 | 5x5 Grid (Mission Control) | Interactive matrix — click cell, fill it | See everything that matters on one page | Job 1, 2 | Prototype |
| 2 | Card View (The Stack) | Swipe through one cell at a time | Daily check-in on mobile in 2 minutes | Job 3 | Gap |
| 3 | Mode Toggle | Switch Pep Talk / Pitch | Same data, different lens | Job 1, 2 | Gap |
| 4 | Scale Selector | Personal / Venture / Industry / Portfolio | Same instrument, different scope | Job 1, 3 | Gap |
| 5 | Save + Persistence | Grid state persists across sessions | Fill today, review next week | All | Gap |
| 6 | Empty Cell Highlight | Visual emphasis on gaps | Gaps become visible strategy | All | Gap |
| 7 | Snapshots | Save dated versions, compare over time | Watch priorities evolve | Job 1, 3 | Gap |
| 8 | Templates | Pre-filled grids for common domains | Start from a pattern, not a blank page | Job 1, 2 | Gap |
| 9 | Share | Public link to read-only grid | Pitch without slides | Job 2 | Gap |
| 10 | AI Coaching (Conversation) | Suggest questions for empty cells | New users get guided, not stuck | All | Gap |
Principles
What truths constrain us?
The Job
| Element | Detail |
|---|---|
| Situation | Someone has too many priorities and no way to see which ones connect |
| Intention | Compress their world into five things that matter — gaps become visible, next move becomes obvious |
| Obstacle | Strategy offsites produce clarity that evaporates. Pitch decks freeze on creation. Journals never compound. |
The hidden objection: "What if I pick the wrong five things?" The grid answers that — wrong picks surface fast because the rows won't fill. The instrument corrects itself.
Why Now
Three forces converge:
- AI makes compression possible — an agent can help fill cells, suggest questions for gaps
- Static pitch decks are dying — investors, teams, and founders want living instruments, not frozen PDFs
- The pattern is proven — every industry analysis, venture scorecard, and PRD on this site is an instance of the same 5x5 grid
The prototype exists. Five interaction patterns explored in CDD. The mental model site IS the filled grid at portfolio scale — 1,200+ pages structured on the same 5P pattern.
Design Constraints
| Constraint | Rationale |
|---|---|
| 5x5 maximum | Cognitive limit — compression IS the value |
| Evergreen beats one-shot | Instruments compound, documents decay |
| Gaps are the strategy | Empty cells are features, not bugs |
Performance
How do we know it works?
Priority Score
PRIORITY = Pain x Demand x Edge x Trend x Conversion
| Dimension | Score (1-5) | Evidence |
|---|---|---|
| Pain | 4 | Every founder struggles with alignment. Strategy offsites produce clarity that evaporates. |
| Demand | 3 | Used internally across 7 ventures. External validation not yet started. |
| Edge | 4 | No tool treats a pitch deck as a living instrument. The 5P constraint is the moat. |
| Trend | 5 | AI makes personal instruments viable. Static documents are dying. |
| Conversion | 2 | No external users yet. Freemium model unvalidated. |
| Composite | 480 | 4 x 3 x 4 x 5 x 2 |
Quality Targets
| # | Criterion | Target | Now |
|---|---|---|---|
| O1 | External users fill a grid | 10 users | 0 |
| O2 | D7 return rate | > 35% | — |
| O3 | Grid completion rate | > 60% of cells filled | — |
| O4 | Shared grids | 3 grids shared externally | 0 |
| O5 | Time to first fill | < 20 minutes | — |
Kill signal: 10 external users fill a grid. If zero return within 7 days, it's a novelty, not an instrument. Stop building and revisit the job. Kill date: 2026-06-30.
Platform
What do we control?
Current State
| Component | Built | Wired | Working | Notes |
|---|---|---|---|---|
| CDD Prototype (5 patterns) | Yes | Yes | Yes | Live |
| Figma Template | Yes | — | — | Visual design + grid layout |
| Stackmates Auth | Yes | Yes | Yes | User accounts, sessions |
| Convex data layer | Yes | Yes | Partial | Real-time sync available |
| Mental model content (1,200+ pages) | Yes | — | — | Templates for grid cells |
System Design
INTERACTION LAYER: Mission Control | Stack | Morning | Compass | AI
DECK ENGINE: 5x5 Matrix | Modes | Scales | Snapshots
DATA LAYER: Convex (Cells | Priorities | Versions | User State)
PLATFORM: Stackmates (Auth | Storage | Sharing | AI Prompts)
Build Ratio
~60% composition (prototype, auth, data layer), ~40% new code (persistence, modes, sharing).
Protocols
How do we coordinate?
Build Order
| Sprint | Features | What | Effort | Acceptance |
|---|---|---|---|---|
| 0 | #1, #5 | Grid component + Convex persistence | 2 weeks | 25 cells render, edit persists across sessions |
| 1 | #2, #3, #4, #6 | Card View + Mode Toggle + Scale + Highlights | 2 weeks | 375px swipe, toggle preserves data, gaps glow |
| 2 | #7, #8, #9 | Snapshots + Templates + Share | 2 weeks | Save snapshot, pre-fill from template, public link works |
| 3 | #10 | AI Coaching | TBD | V3 — after external validation |
Commissioning
| # | Feature | Install | Test | Operational | Optimize |
|---|---|---|---|---|---|
| 1 | 5x5 Grid | Prototype | — | — | — |
| 2 | Card View | — | — | — | — |
| 3 | Mode Toggle | — | — | — | — |
| 4 | Scale Selector | — | — | — | — |
| 5 | Save + Persistence | — | — | — | — |
| 6 | Empty Cell Highlight | — | — | — | — |
| 7 | Snapshots | — | — | — | — |
| 8 | Templates | — | — | — | — |
| 9 | Share | — | — | — | — |
| 10 | AI Coaching | — | — | — | — |
Players
Who creates harmony?
Job 1: Founder Alignment
Situation: Starting a new venture or entering a new quarter with too many priorities.
| Element | Detail |
|---|---|
| Struggling moment | Strategy session produces 30 items. Can't see which five matter. |
| Current workaround | Notion board, Google Doc, whiteboard photo that fades |
| What progress looks like | 25 cells filled. Gaps visible. Weekly review takes 10 minutes. |
| Hidden objection | "I don't want to commit to five — what if I'm wrong?" |
| Switch trigger | Third strategy session in a row that produces no lasting artifact |
Features that serve this job: #1, #3, #5, #7
Job 2: Living Pitch
Situation: Preparing to pitch an idea to investors, partners, or a team.
| Element | Detail |
|---|---|
| Struggling moment | Pitch deck freezes at version 1. Reality moves. Slides don't. |
| Current workaround | Remake slides before every meeting. Copy-paste from old decks. |
| What progress looks like | One living deck that compounds — each review sharpens the pitch |
| Hidden objection | "Investors expect slides, not a grid" |
| Switch trigger | Pitched with stale data. Lost credibility. |
Features that serve this job: #1, #3, #8, #9
Job 3: Personal Navigation
Situation: Life transition — career change, health goal, relationship rebuild.
| Element | Detail |
|---|---|
| Struggling moment | Too many dimensions changing at once. No single view. |
| Current workaround | Journal, therapy, self-help books — all separate, none connected |
| What progress looks like | Five life priorities visible. Gaps named. Weekly check-in ritual. |
| Hidden objection | "Self-reflection tools feel cringe" |
| Switch trigger | Something breaks — burnout, breakup, wake-up call |
Features that serve this job: #2, #4, #7
ICP: The Misaligned Operator
Founders, team leads, and operators who run strategy sessions that produce clarity — then watch it evaporate. They don't need another canvas. They need a constraint that makes compression deliberate.
Psycho-logic: "I know I should prioritise. I just can't see it all on one page." The hidden fear isn't picking wrong — it's that seeing the gaps will reveal how much they don't know. The grid makes that safe by normalising gaps as strategy, not failure.
Relationship to Other PRDs
| PRD | Relationship | Data Flow |
|---|---|---|
| Time + Mind | Peer | Prompt Deck aligns WHAT matters → Time+Mind closes gap between WHAT and DID |
| Agent Platform | Platform (depends on) | Auth, AI agent paths, data layer |
| Sales Dev Agent | Peer | Filled grid feeds ICP qualification |
Context
- Pictures — Pre-flight maps that feed this spec
- Prompt Deck — Sales compression of this spec
- Tight Five Platform — The pattern the product implements
- Live Prototype — Current CDD build
- ICP Framework — Target definition with psycho-logic
- AI Product Requirements — Section definitions and checklist