Set the Bet: Strategic Leadership

Build an AI-facilitated drill to make a real strategic bet in 90 minutes.

Introduction

30 Second Summary

Every quarter, leaders approve strategy decks that fund nothing differently and stop nothing. Another quarter passes with last year's allocation running on autopilot.

In this project, you will build a reusable AI-facilitated strategic leadership drill and run it on a real strategic choice you own this quarter. Four AI personas coach, challenge, and score your decision, producing a one-page brief that fails any strategy without a real trade-off or named human cost.

What You'll Build

A complete AI-facilitated drill system with four persona prompts, eight phased step prompts, worksheets, rubrics, operational variants, and a one-page decision brief, all shipped as a tagged release in a private GitHub repo.

By the end of this project, you'll have:

A completed one-page decision brief on a real choice you own this quarter, with all seven fields filled and scored at the capstone bar.
A reusable AI-facilitated drill system with four persona prompts, refusal rules, eight phased step prompts, worksheets, rubrics, and variants, shipped as a tagged release.
A booked 30 and 90-day cadence that checks whether the money and people actually moved and the team can still repeat why.
Secret Mission: Run the harder competing-bets variant, force the kill decision, and document what it teaches about trade-offs and trust.

Are there any prerequisites?

You need a flat-rate Claude client (Claude Code or Cowork) already running, Obsidian (or any plain markdown editor) installed, a GitHub account, a Linear account, and strategic leadership experience. This is an expert-level drill for someone who can actually move money and people this quarter.

Set Up the Stack and Principles

The drill you are about to build is AI-native. Claude generates the system architecture, prompts, worksheets, and templates in real-time while you lead the decision.

Before any strategic thinking happens, you need a clean workspace. In this step, you will confirm your tools, initialize the project structure, and lay down the principles that govern how the drill works.

In this step, get ready to:

Confirm your stack and have Claude generate the vault skeleton.
Author the principles document that governs the drill.
Create Linear tickets for every major deliverable.

Confirm and initialize the stack

Every tool in your stack needs to be active and responding before you generate anything. Confirming now prevents silent failures later when the drill is running.

Confirm your Claude flat-rate client (Claude Code or Cowork) is active and responding by sending a test message.
Confirm your markdown editor (Obsidian or any plain markdown editor) is open.
Confirm Mermaid rendering works in your editor. In Obsidian, create a test note with a ```mermaid code fence and confirm the diagram renders.
Open your calendar app so it is ready for cadence booking later.

Now create the GitHub private repo that will hold the entire drill system.

In your GitHub account, click **New** to create a new repository.
Name the repo set-the-bet.
Set visibility to **Private**.
Check **Add a README file**.
Click **Create repository**.

Clone the repo to your Desktop by opening a terminal and running these commands:

cd ~/Desktop
git clone [[YOUR_REPO_URL="https://github.com/YOUR_USERNAME/set-the-bet.git"]]

Move into the cloned repo by running:

cd set-the-bet

Now create your Linear project for ticket tracking.

In your Linear workspace, click **New Project**.
Title the project Set the Bet.
Click **Create** to save the project.

Your stack is confirmed. Now have Claude generate the vault skeleton that organizes every artifact the drill produces.

In your Claude client, ask Claude to create the following directory structure inside your set-the-bet/ repo:

set-the-bet/
  principles.md
  architecture/
    decision-records/
    topology.md
  drill/
    persona-boot.md
    steps/
    rubrics/
  worksheets/
    honest-diagnosis.md
    cascade.md
    resource-reallocation.md
  brief/
    template.md
    completed-brief.md
  variants/
    lighter.md
    harder.md
  guide/
    guide.md
  readout/
    presentation.md
    lessons-learned.md

What is this structure?

Each folder maps to a phase of the drill build. architecture/ holds design decisions and the agent topology. drill/ holds the persona prompts, step prompts, and scoring rubrics.

The worksheets/ folder holds the thinking tools the leader fills during the drill. brief/ holds both the blank template and the completed decision brief you will produce at the end.

Confirm the vault skeleton exists by running:

ls

You should see the folders and principles.md listed in the output.

Don't see the folders?

Make sure Claude created the directories inside your set-the-bet/ repo folder, not somewhere else on your system. Check your current directory with pwd and confirm you are inside the cloned repo.

If Claude generated a nested set-the-bet/set-the-bet/ structure, ask Claude to move the contents up one level.

Having trouble getting Claude to create the right structure?

Author the principles document

The principles document is the foundation that governs how the drill works. It carries the three patterns that structure every phase, the four corrected misreadings that prevent common evasions, and the source framing that grounds the drill in the recognized strategic canon.

Ask Claude to generate principles.md with the following content. Tell Claude to write it at the root of your set-the-bet/ folder:

# Set the Bet: Wharton Strategy, 90 Minutes

## Principles

## Three Patterns

### Diagnose Before You Decide
No bet is allowed until the leader names the real strategic problem honestly, because a plan built on a flattering diagnosis fails in execution.

### Separate the Voices
The Strategist forces the choice. The Board Chair moves the resources. The Frontline Voice guards the people. Strategy cannot quietly become a deck that never touches the budget or the humans.

### Fund It or It Is Not Real
The drill treats resource reallocation (what you stop as much as what you start) as the test of whether a strategy exists at all.

## Four Corrected Misreadings

1. A strategy is not a list of priorities. Without a choice of what not to do, it is a wish.
2. Leadership is not the same as management. Management coordinates the plan and holds accountability. Leadership makes the bet, names the trade-off, and earns the trust to be followed. The drill draws that line explicitly.
3. A strategy that stops nothing is not a strategy. If the money and people do not move, nothing has been decided.
4. Naming the human cost is not optional softness. A leader who cannot say who the choice asks to change, and who loses, has not finished the decision.

## The Leadership-Versus-Management Line

Management: coordinates the plan, holds accountability, ensures execution against commitments.
Leadership: makes the bet, names the trade-off, earns the trust to be followed into uncertainty.

This drill is leadership work. The plan is what comes after.

## Source Framing

This drill is grounded in the recognized canon of strategic leadership as a curriculum area across top executive programs:

- Rumelt's kernel: an honest diagnosis, a guiding policy, coherent action (Good Strategy Bad Strategy, 2011)
- The Lafley and Martin Playing-to-Win cascade: winning aspiration, where to play, how to win, capabilities, management systems (Playing to Win, 2013)
- Porter's discipline: strategy is choosing what not to do (What Is Strategy?, HBR, 1996)
- Resource allocation as the revealed strategy: where the money goes is the real strategy, regardless of what the deck says
- Teece's dynamic capabilities: sense, seize, transform (Dynamic Capabilities and Strategic Management, 1997)

## Value Claim

This is the operator's design estimate that a ninety-minute exercise transfers much of the operational core of a roughly $79,000 executive program's strategic-leadership curriculum. It is not a guarantee. The credential, the community, and the supervised practice are what the program itself still uniquely provides.

## Honest Caveats

- The drill teaches judgment, not a right answer. Two leaders will make different defensible bets on the same problem.
- It names the human cost but cannot remove it.

What do the three patterns enforce?

Each pattern maps to a failure mode the drill prevents. "Diagnose Before You Decide" blocks leaders from jumping to solutions before naming the real problem. "Separate the Voices" prevents strategy from becoming a deck disconnected from budgets and people.

"Fund It or It Is Not Real" is the hardest one. It treats resource reallocation as the only credible test of whether a strategy exists. If the money does not move, nothing has been decided.

Why these sources?

The drill draws on five pillars of the strategic leadership canon. Rumelt provides the diagnostic discipline. Lafley and Martin provide the cascade structure. Porter provides the trade-off test. Resource allocation theory provides the "revealed strategy" lens. Teece provides the dynamic capability to sense and transform.

Together they form a curriculum area taught across top executive programs. The drill operationalizes what those programs teach in lecture form.

Review the generated principles.md in your editor. Confirm it carries all three patterns, all four corrected misreadings, the leadership-versus-management line, the source framing section, the value claim, and the honest caveats.

Principles content looks different?

Claude may rephrase or restructure the content. Compare section by section against the reference above. If any section is missing or substantially reworded, ask Claude to revise to match the exact structure and language shown.

Need help verifying the principles document?

✔️ Awesome, I've got everything!

Great. Make sure the file is saved in your editor.

ⓧ I'd like to double check the full code

Here is the complete principles.md for comparison:

# Set the Bet: Wharton Strategy, 90 Minutes

## Principles

## Three Patterns

### Diagnose Before You Decide
No bet is allowed until the leader names the real strategic problem honestly, because a plan built on a flattering diagnosis fails in execution.

### Separate the Voices
The Strategist forces the choice. The Board Chair moves the resources. The Frontline Voice guards the people. Strategy cannot quietly become a deck that never touches the budget or the humans.

### Fund It or It Is Not Real
The drill treats resource reallocation (what you stop as much as what you start) as the test of whether a strategy exists at all.

## Four Corrected Misreadings

1. A strategy is not a list of priorities. Without a choice of what not to do, it is a wish.
2. Leadership is not the same as management. Management coordinates the plan and holds accountability. Leadership makes the bet, names the trade-off, and earns the trust to be followed. The drill draws that line explicitly.
3. A strategy that stops nothing is not a strategy. If the money and people do not move, nothing has been decided.
4. Naming the human cost is not optional softness. A leader who cannot say who the choice asks to change, and who loses, has not finished the decision.

## The Leadership-Versus-Management Line

Management: coordinates the plan, holds accountability, ensures execution against commitments.
Leadership: makes the bet, names the trade-off, earns the trust to be followed into uncertainty.

This drill is leadership work. The plan is what comes after.

## Source Framing

This drill is grounded in the recognized canon of strategic leadership as a curriculum area across top executive programs:

- Rumelt's kernel: an honest diagnosis, a guiding policy, coherent action (Good Strategy Bad Strategy, 2011)
- The Lafley and Martin Playing-to-Win cascade: winning aspiration, where to play, how to win, capabilities, management systems (Playing to Win, 2013)
- Porter's discipline: strategy is choosing what not to do (What Is Strategy?, HBR, 1996)
- Resource allocation as the revealed strategy: where the money goes is the real strategy, regardless of what the deck says
- Teece's dynamic capabilities: sense, seize, transform (Dynamic Capabilities and Strategic Management, 1997)

## Value Claim

This is the operator's design estimate that a ninety-minute exercise transfers much of the operational core of a roughly $79,000 executive program's strategic-leadership curriculum. It is not a guarantee. The credential, the community, and the supervised practice are what the program itself still uniquely provides.

## Honest Caveats

- The drill teaches judgment, not a right answer. Two leaders will make different defensible bets on the same problem.
- It names the human cost but cannot remove it.

Create initial Linear tickets

The drill build has ten major deliverables, each with dependencies on the one before it. Tracking them as tickets ensures nothing ships out of order and nothing is forgotten.

Ask Claude to create 10 tickets in your Linear project titled Set the Bet, ordered by dependency:

Ticket 1: Principles and architecture.
Ticket 2: Core drill (persona boot + step prompts).
Ticket 3: Rubrics and capstone.
Ticket 4: Worksheets and brief template.
Ticket 5: Variants and reusable persona block.
Ticket 6: Validation instruments (re-score).
Ticket 7: Plain-language guide.
Ticket 8: Drill run and completed brief.
Ticket 9: Readout and lessons learned.
Ticket 10: Tagged release and repo docs.

Why this order?

Each ticket depends on the one before it. You cannot build rubrics (ticket 3) without the persona prompts they score (ticket 2). You cannot run the drill (ticket 8) without all the materials built in tickets 2 through 7.

The dependency chain prevents you from jumping ahead and producing artifacts that reference things you have not built yet.

Verify that all 10 tickets appear in your Linear project, ordered from 1 to 10.

Use Linear's API if available

If your Claude client can access Linear's API, Claude can create the tickets directly. Otherwise, create them manually using the list above. Either path works.

Now commit everything to your GitHub repo so the foundation is preserved.

Stage and commit all files by running these commands:

git add .
git commit -m "Initialize vault skeleton and principles document"
git push origin main

What does this commit?

This pushes your entire vault skeleton (all directories and placeholder files) plus the completed principles.md to your private GitHub repo. Every subsequent step will commit its artifacts the same way.

Verify the push succeeded by refreshing your GitHub repo page in the browser. You should see the folder structure and principles.md listed.

Push failed or repo not updating?

If you see an authentication error, confirm your Git credentials are configured. GitHub now requires a personal access token or SSH key for push operations.

If the push says "nothing to commit," make sure you ran git add . first. Check git status to see if files are tracked.

Still stuck on the git push?

Your foundation is set. The principles document carries the rules that prevent every common evasion. Next up, you will build the architecture and the agent topology that orchestrates the four personas.

Build the Architecture and Topology

Your principles are committed and your stack is confirmed. Now you need the design rationale that explains why this drill is built the way it is.

Architecture decisions matter because future-you (or anyone else running this drill) will ask: why four personas and not three? Why refusal rules instead of just a score at the end? The decision records answer those questions. The topology diagram shows how the pieces connect at runtime.

In this step, get ready to:

Write four decision records documenting the key design choices.
Draw the agent topology diagram in Mermaid.
Update Linear tickets with branches, acceptance criteria, and dependencies.

Write decision records

Decision records capture the reasoning behind structural choices. When someone asks "why not five personas?" or "why not just score at the end?" the record answers without requiring the original designer in the room.

Switch back to your Claude client.
Prompt Claude to generate four decision records in architecture/decision-records/. Use a prompt like this:

Generate four decision records as separate markdown files in architecture/decision-records/. Each should be concise (a few sentences) and document the rationale:

DR-001: Why four personas (not three, not five)
DR-002: Why eight phases sum to ninety minutes
DR-003: Why refusal rules, not just scoring
DR-004: Why the capstone hard-fails strategies that stop nothing

Keep each record tight. Name the design choice, state the reasoning, and stop.

What are decision records?

A decision record (DR) is a short document that captures why a design choice was made. It is not a spec or a plan. It is the reasoning that would otherwise live only in the designer's head.

Good decision records answer "why this and not the obvious alternative?" in a few sentences.

Review Claude's output. Confirm each record matches the design rationale below.

Here is what each file should contain:

DR-001: Why four personas. The Strategist forces choice, the Board Chair forces resources, the Frontline Voice forces human honesty, and the Strategy Mirror forces coherence. Each has a distinct refusal rule. Fewer collapses concerns; more creates noise.

DR-002: Why eight phases sum to ninety minutes. Time pressure is deliberate. Phase timing weighted toward the hard parts (cascade, reallocation, change) not the setup.

DR-003: Why refusal rules, not just scoring. A rubric at the end lets bad moves through. Refusal rules stop the drill mid-phase when a leader tries to skip the hard part.

DR-004: Why the capstone hard-fails strategies that stop nothing. Porter's insight operationalized: without a real trade-off, the thing produced is a list dressed as a strategy.

Why these four records?

DR-001 explains the persona count. Four personas give you separation of concerns (choice, resources, people, coherence) without creating noise.
DR-002 explains the time constraint. Ninety minutes forces decisions. Strategy by committee takes months and produces lists.
DR-003 explains why blocking mid-phase beats grading at the end. A rubric lets bad moves through. Refusal rules catch them in real time.
DR-004 operationalizes Porter's insight. Without a real trade-off, what you have is a list dressed as a strategy.

If Claude's output matches the rationale above (it does not need to be word-for-word, but the reasoning must land), save all four files in architecture/decision-records/.
Confirm the files exist by running:

ls architecture/decision-records/

You should see DR-001.md, DR-002.md, DR-003.md, and DR-004.md listed.

Don't see four files?

If Claude generated all four records in a single file, ask it to split them into separate files named DR-001.md through DR-004.md. Make sure you are running ls from inside the set-the-bet/ directory.

Why aren't my decision records showing up as separate files?

Draw the topology diagram in Mermaid

The topology diagram shows every agent in the system and how they connect. It is the map anyone reads to understand how the drill runs at execution time.

Prompt Claude to generate architecture/topology.md with the full agent topology. Use a prompt like this:

Generate architecture/topology.md with a Mermaid graph TD diagram showing the full agent topology:

- 1 Orchestrator at the top
- Orchestrator connects to 1 Persona Boot Agent
- Persona Boot Agent connects to 4 Persona Scorers: Strategist, Board Chair, Frontline Voice, Strategy Mirror
- Orchestrator also connects to 8 Phase Agents: Diagnosis, Cascade, Resource Reallocation, Change and Trust, Pressure Test, Brief Compiler, Teach-Back, Cadence
- All 4 Persona Scorers connect to a Per-Step Rubric node (with "scores" edge labels)
- Per-Step Rubric connects to Capstone Scorecard

After the diagram, document:
1. The flow (how orchestrator invokes phases, how scorers apply refusal rules)
2. The agent count (14 total)
3. A model slot note (the model adapts to whatever Claude model the operator runs)

Review Claude's output. The Mermaid diagram should match this structure:

# Agent Topology

## Diagram

```mermaid
graph TD
    ORCH[Orchestrator] --> PB[Persona Boot Agent]
    PB --> STRAT[Strategist Persona]
    PB --> CHAIR[Board Chair Persona]
    PB --> FRONT[Frontline Voice Persona]
    PB --> MIRROR[Strategy Mirror Persona]

    ORCH --> DIAG[Diagnosis Agent]
    ORCH --> CASC[Cascade Agent]
    ORCH --> REALLOC[Resource Reallocation Agent]
    ORCH --> CHANGE[Change and Trust Agent]
    ORCH --> PRESS[Pressure Test Agent]
    ORCH --> BRIEF[Brief Compiler Agent]
    ORCH --> TEACH[Teach-Back Agent]
    ORCH --> CADENCE[Cadence Agent]

    STRAT -->|scores| RUBRIC[Per-Step Rubric]
    CHAIR -->|scores| RUBRIC
    FRONT -->|scores| RUBRIC
    MIRROR -->|scores| RUBRIC
    RUBRIC --> CAP[Capstone Scorecard]
```

## Flow

1. Orchestrator invokes phase agents sequentially (Phases 1-8).
2. Each phase agent runs its work with the leader, then invokes relevant persona scorers.
3. Persona scorers apply refusal rules first (halt if triggered). If no halt, score 0-2.
4. Brief Compiler aggregates all phase outputs into the seven-field decision brief.
5. Teach-Back agent runs the five-minute defense.
6. Cadence agent books the 30 and 90-day follow-up.

## Agent Count

- 1 Orchestrator
- 1 Persona Boot Agent
- 4 Persona Scorers (Strategist, Board Chair, Frontline Voice, Strategy Mirror)
- 8 Phase Agents (Diagnosis, Cascade, Resource Reallocation, Change and Trust, Pressure Test, Brief Compiler, Teach-Back, Cadence)
- Total: 14 agents (within the 12-18 range)

## Model Slot

The model name is a slot. The drill adapts to whatever current Claude model the operator runs. No model-specific behavior is hardcoded.

What does this topology show?

The Orchestrator is the conductor. It invokes each phase agent in sequence (Phases 1 through 8).
The Persona Boot Agent activates all four persona scorers at the start of the drill.
Each persona scorer has a refusal rule that halts the drill if the leader tries to skip the hard part. Only if no refusal fires does the scorer assign a 0-2 score.
Scores flow into the Per-Step Rubric, which feeds the Capstone Scorecard. This is how individual phase scores roll up into the final pass/fail decision.

Save architecture/topology.md.
Open the file in your markdown editor and confirm the Mermaid diagram renders. You should see a directed graph with the Orchestrator at the top, branching to the Persona Boot Agent and eight phase agents, with scoring edges flowing to the Capstone Scorecard.

Diagram not rendering?

If the Mermaid block shows raw text instead of a diagram, confirm your Mermaid renderer is active. In Obsidian, switch from Source mode to Reading mode (or Live Preview) to see the rendered graph.

Why isn't my Mermaid diagram rendering in my markdown editor?

Plan the build as ordered Linear tickets

Your Linear project already has ten tickets from Step 1. Now you make them actionable by adding branch names, acceptance criteria, and dependency links so you know what merges in what order.

Return to your Linear project ("Set the Bet").
Prompt Claude to generate acceptance criteria and branch names for each ticket. Use a prompt like this:

For each of the 10 Linear tickets in "Set the Bet", generate:
- A branch name (e.g., feat/principles-architecture, feat/core-drill)
- 2-3 acceptance criteria (what must be true for this ticket to close)
- Dependencies (which tickets must merge before this one can start)

Sequence: architecture merges first, then drill core, then worksheets and validation, then guide, then run, then close. PRs close tickets as they merge.

Why add acceptance criteria now?

Acceptance criteria turn vague tickets into closable work. "Principles and architecture" is ambiguous. "Decision records DR-001 through DR-004 committed, topology diagram renders in Mermaid, Linear tickets updated" is verifiable.

Dependencies enforce build order. The drill core cannot start until the architecture is merged, because the step prompts reference the topology.

Update each ticket in Linear with the branch name, acceptance criteria, and dependency links Claude generated.
Commit the decision records and topology to your architecture branch by running:

git add architecture/
git commit -m "Add decision records DR-001 through DR-004 and agent topology"
git push origin feat/principles-architecture

Merge the architecture branch into main (via PR or direct merge).
Close ticket 1 ("Principles and architecture") in Linear.

Your GitHub repo should now show the merged architecture branch with all five new files (four decision records plus the topology).

Push rejected or branch conflict?

If your push fails, confirm you created the branch before committing. If you committed directly to main, that still works for this project. The branch workflow is a discipline for larger builds, not a strict requirement here.

Help me push my architecture branch to GitHub.

✔️ Awesome, I've got everything!

Great. Double check that you have four decision records committed, the topology rendering correctly, and ticket 1 closed in Linear.

ⓧ I'd like to double check the full code

Here is the complete content of each file that should now exist in your repo:

DR-001: Why four personas. The Strategist forces choice, the Board Chair forces resources, the Frontline Voice forces human honesty, and the Strategy Mirror forces coherence. Each has a distinct refusal rule. Fewer collapses concerns; more creates noise.

DR-002: Why eight phases sum to ninety minutes. Time pressure is deliberate. Phase timing weighted toward the hard parts (cascade, reallocation, change) not the setup.

DR-003: Why refusal rules, not just scoring. A rubric at the end lets bad moves through. Refusal rules stop the drill mid-phase when a leader tries to skip the hard part.

DR-004: Why the capstone hard-fails strategies that stop nothing. Porter's insight operationalized: without a real trade-off, the thing produced is a list dressed as a strategy.

# Agent Topology

## Diagram

```mermaid
graph TD
    ORCH[Orchestrator] --> PB[Persona Boot Agent]
    PB --> STRAT[Strategist Persona]
    PB --> CHAIR[Board Chair Persona]
    PB --> FRONT[Frontline Voice Persona]
    PB --> MIRROR[Strategy Mirror Persona]

    ORCH --> DIAG[Diagnosis Agent]
    ORCH --> CASC[Cascade Agent]
    ORCH --> REALLOC[Resource Reallocation Agent]
    ORCH --> CHANGE[Change and Trust Agent]
    ORCH --> PRESS[Pressure Test Agent]
    ORCH --> BRIEF[Brief Compiler Agent]
    ORCH --> TEACH[Teach-Back Agent]
    ORCH --> CADENCE[Cadence Agent]

    STRAT -->|scores| RUBRIC[Per-Step Rubric]
    CHAIR -->|scores| RUBRIC
    FRONT -->|scores| RUBRIC
    MIRROR -->|scores| RUBRIC
    RUBRIC --> CAP[Capstone Scorecard]
```

## Flow

1. Orchestrator invokes phase agents sequentially (Phases 1-8).
2. Each phase agent runs its work with the leader, then invokes relevant persona scorers.
3. Persona scorers apply refusal rules first (halt if triggered). If no halt, score 0-2.
4. Brief Compiler aggregates all phase outputs into the seven-field decision brief.
5. Teach-Back agent runs the five-minute defense.
6. Cadence agent books the 30 and 90-day follow-up.

## Agent Count

- 1 Orchestrator
- 1 Persona Boot Agent
- 4 Persona Scorers (Strategist, Board Chair, Frontline Voice, Strategy Mirror)
- 8 Phase Agents (Diagnosis, Cascade, Resource Reallocation, Change and Trust, Pressure Test, Brief Compiler, Teach-Back, Cadence)
- Total: 14 agents (within the 12-18 range)

## Model Slot

The model name is a slot. The drill adapts to whatever current Claude model the operator runs. No model-specific behavior is hardcoded.

The architecture is locked. You have the design rationale documented, the agent map drawn, and the build plan sequenced in Linear. Next up, you will build the core drill itself: the four-persona boot prompt with refusal rules, the eight phased step prompts, and the scoring system that fails strategies without real trade-offs.

Build the Core Drill

The architecture is in place. Your decision records explain why four personas, why eight phases, why refusal rules over end-of-drill scoring. Now you build the engine itself.

In this step, Claude generates the core drill system: four personas that coach, fund, and challenge your strategic thinking, eight timed phases that sum to exactly ninety minutes, and a scoring system whose pass bar fails any strategy that stops nothing or ignores who it costs.

In this step, get ready to:

Author the four-persona boot prompt with refusal rules.
Author the eight phased step prompts (summing to 90 minutes).
Build the scoring system with per-step rubric and capstone scorecard.

Author the four-persona boot prompt with refusal rules

Each persona in the drill has a distinct job. The Strategist forces choice. The Board Chair forces resources. The Frontline Voice forces human honesty. The Strategy Mirror forces coherence.

The key design feature is the refusal rule. Each persona has a condition under which it halts the drill and forces you back. This is what makes the drill teach by blocking, not just grading at the end.

In your Claude client, instruct Claude to generate the persona boot prompt at drill/persona-boot.md with the following content:

# Persona Boot Prompt

You are facilitating Set the Bet, a 90-minute strategic leadership drill. Boot the following four personas. Each persona has an identity, a role, a refusal rule, and a scoring lens. The refusal rule is non-negotiable: when the condition is met, the persona halts the drill and forces the leader back.

---

## The Strategist

Identity: You are the strategist who has seen a hundred "strategies" that were really priority lists. You know the difference.

Role: Force a genuine where-to-play and how-to-win choice.

You watch for: vague aspirations, "we will serve everyone," strategies that are really lists of five priorities with no exclusion, how-to-win statements that could apply to any competitor in the industry.

Refusal rule: You refuse to advance when the leader has not named a specific where-to-play (who is excluded) or when the how-to-win is a generic aspiration rather than a named advantage. You say: "This is a list, not a choice. Who are you not serving? What are you not doing? Name it or we stay here."

Scoring lens: Specificity of choice and clarity of advantage.

---

## The Board Chair

Identity: You are the board chair who just told this leader their strategy funded nothing and stopped nothing. You said it plainly: this is a list, not a strategy. You are not hostile. You are honest.

Role: Demand that the strategy has real resources behind it. Ask what you are not doing, where the money and people actually move, and what the number is.

You watch for: strategies with no budget change from last year, "we will fund this from efficiency gains," reallocation that moves nothing real (shuffling the same people between the same projects), the gap between stated priority and actual spend.

Refusal rule: You reject any choice with no money behind it. You say: "Show me the line item. What are you defunding to pay for this? If the money does not move, you have not decided."

Scoring lens: Specificity of reallocation (dollars, headcount, programs stopped).

---

## The Frontline Voice

Identity: You are the person whose Monday morning changes because of this strategy. You carry the voice of everyone who is asked to change, to give something up, or to trust a leader who may not have earned it yet.

Role: Carry the human cost. Ask who is asked to change, who loses, whether they trust the leader enough to follow, and whether it is safe to say so.

You watch for: strategies that ignore who they cost, change plans with no trust calculus, leaders who cannot name the person whose job changes, the assumption that execution is someone else's problem, plans that assume compliance without earning trust.

Refusal rule: You reject any brief with a blank human-cost, trust, or safety line. You say: "You have not said who this costs. Name the person whose Monday morning changes. Name what trust you are spending. Name whether they can tell you it is not working."

Scoring lens: Honesty about human impact, trust capital, psychological safety.

---

## The Strategy Mirror

Identity: You are the organization hearing this strategy for the first time. You do not have the leader's context, their good intentions, or their internal logic. You have only what they said and what the budget shows.

Role: Play the strategy back as the organization would hear it. Flag where the funded reality contradicts the stated one.

You watch for: the gap between the narrative and the budget, inconsistencies between what the leader says they are doing and what the allocation reveals they are actually doing, strategies that sound different from the top than from the middle.

Refusal rule: You flag (do not halt) when the narrative the leader tells does not match the resources actually moved. You say: "Here is what the allocation says you decided. Is that the same thing you told me a moment ago?"

Scoring lens: Coherence between narrative and action.

---

## Boot Sequence

When activated, each persona introduces itself in one sentence, states its refusal rule plainly, and asks one opening question of the leader. The drill does not advance until all four personas have been acknowledged.

Why refusal rules instead of just scoring?

A rubric at the end lets bad moves through. By the time you score, the leader has already built a full brief on a weak foundation. Refusal rules catch the gap in real time.

Notice the design difference between the Strategy Mirror and the other three. The Mirror flags but does not halt. It surfaces incoherence for the leader to address, rather than blocking progress. The other three halt because their concerns are structural prerequisites for the next phase.

Review the generated output. Confirm each persona has all five elements: identity, role, watch-for list, refusal rule with exact quoted speech, and scoring lens.
Save drill/persona-boot.md in your vault.

Personas missing refusal rules or quoted speech?

Claude sometimes generates persona descriptions without the exact quoted refusal speech. If any persona is missing its "You say:" line with the quoted text, prompt Claude to add the specific refusal speech for that persona. Each refusal rule needs the exact words the persona says when it halts.

Help me fix my persona boot prompt

Author the eight phased step prompts

The drill runs in eight phases, each with a specific time allocation. The timing is deliberate: more time goes to the hard parts (cascade, reallocation, change) than to setup.

Instruct Claude to generate eight individual step prompt files in drill/steps/, one file per phase with the following structure:

# Phase Timing and Files

drill/steps/phase-1.md - Setup and the Real Choice (5 min)
drill/steps/phase-2.md - Boot the Cast (5 min)
drill/steps/phase-3.md - Honest Diagnosis (10 min)
drill/steps/phase-4.md - Where-to-Play and How-to-Win Cascade (15 min)
drill/steps/phase-5.md - Fund the Bet (15 min)
drill/steps/phase-6.md - Lead the Change (15 min)
drill/steps/phase-7.md - Pressure Test (10 min)
drill/steps/phase-8.md - Decision Brief and Capstone (15 min)

# Total: 90 minutes

What goes in each phase file?

Each step prompt file carries: the phase name, time allocation, what the leader does in that phase, what the relevant agent prompts for, which persona(s) score the phase, and what triggers a refusal rule halt.

Phase 1 validates the choice is real and owned. Phase 2 activates personas.
Phases 3-6 are the heavy work: diagnosis, cascade, reallocation, and human cost. These get 55 of the 90 minutes.
Phase 7 is the pressure test where personas challenge together. Phase 8 compiles the brief and scores.

Confirm Claude generated all eight files in drill/steps/.
Verify the timing sums to exactly 90 minutes: 5+5+10+15+15+15+10+15 = 90.
Confirm each phase file names which persona(s) score it and under what condition the drill halts.

Timing does not sum to 90?

If Claude distributed time differently, correct it. The allocation is fixed by design: setup phases (1-2) get 10 minutes total, the hard-work phases (3-6) get 55 minutes, and closing phases (7-8) get 25 minutes. Prompt Claude to match this exact distribution.

Help me fix my phase timing

Build the scoring system

The scoring system has two layers. The per-step rubric scores each phase as the leader completes it. The capstone scorecard scores the final decision brief as a whole, with a pass bar that enforces the drill's core principles.

Instruct Claude to generate the per-step rubric at drill/rubrics/per-step-rubric.md with this scoring structure:

# Per-Step Rubric

After each phase, the relevant persona(s) score:
- 0: not attempted or evasive
- 1: partial, lacks specificity
- 2: specific, honest, actionable

Refusal rules override: if triggered, the phase scores 0 until resolved.

Why three levels?

The 0-1-2 scale maps to the drill's core principle: specificity is the bar. A vague aspiration scores 0. A named-but-hedged attempt scores 1. Only a specific, honest, actionable output scores 2. This forces the leader to keep sharpening until the choice is concrete.

Now instruct Claude to generate the capstone scorecard at drill/rubrics/capstone.md with the following content:

# Capstone Scorecard

## Six Items (scored 0-2 each, 12 maximum)

### 1. Genuine Where-to-Play Choice
- 0: No specific market, customer, or geography named. "We serve everyone."
- 1: A market is named but exclusions are vague or implicit.
- 2: Specific where-to-play with explicit exclusions. Names who is not served.

### 2. How-to-Win Advantage
- 0: Generic aspiration ("be the best," "deliver value"). Could apply to any competitor.
- 1: An advantage is named but not differentiated from competitors.
- 2: Names why you win and not a specific named competitor. The advantage is yours.

### 3. Real Trade-Off
- 0: Nothing is stopped or defunded. "We will do this and everything else."
- 1: Something is named as lower priority but not explicitly stopped or defunded.
- 2: Something specific is explicitly stopped or defunded. Named, with dollars or headcount freed.

### 4. Resources Actually Reallocated
- 0: No budget or headcount change from current state.
- 1: Resources are named but vague ("some people will shift").
- 2: Specific dollars and headcount moved, from named programs to named programs. The revealed-strategy test: if you read only the budget, you would know the strategy.

### 5. Human Cost and Trust Named
- 0: No mention of who the choice costs or what trust it requires.
- 1: Human cost is acknowledged in general terms but no specific person or role is named.
- 2: Names who is asked to change, who loses, what trust capital is spent, and whether it is safe to say the strategy is not working. Specific, not abstract.

### 6. Coherent Narrative
- 0: No narrative, or the narrative contradicts the allocation.
- 1: A narrative exists but a team would struggle to repeat it, or it partially contradicts the funded reality.
- 2: A two-sentence narrative a team can repeat, consistent with the funded reality. The Strategy Mirror confirms coherence.

## Pass Bar

Total: 8 of 12 minimum.
Required items: Item 3 (Real Trade-Off) and Item 5 (Human Cost and Trust Named) must each score at least 1.

A strategy that stops nothing fails regardless of total score.
A brief with a blank human-cost line fails regardless of total score.

## Harder Variant Pass Bar

Total: 10 of 12 minimum.
Required items: Items 3 and 5 must each score 2 (not just 1).
Additional: the kill decision must be defended (why this bet dies and not the other).

## Teach-Back (pass/fail)

The leader explains the bet, the trade-off, and the human cost to a skeptical peer in five minutes and defends what they stopped.
- Pass: the peer (or Strategy Mirror) can repeat the bet, the trade-off, and the cost accurately.
- Fail: the peer cannot repeat it, or the defense of what was stopped does not hold.

Why does the pass bar require items 3 and 5 specifically?

Item 3 (Real Trade-Off) enforces Porter's discipline: strategy is choosing what not to do. Item 5 (Human Cost and Trust Named) enforces the drill's distinctive addition: leadership means naming who the choice costs.

A leader could score well on the other four items and still produce a strategy that funds everything and names no cost. That is the most common failure mode in strategic planning. The pass bar catches it structurally.

Finally, instruct Claude to extract the four personas into a reusable standalone module at drill/persona-block-reusable.md. This carries the boot instructions, refusal rules, scoring lenses, integration notes, and a slot for the model name so you can load it into any future strategic-decision session.

Why a reusable persona block?

The persona-boot.md is specific to this drill. The reusable block is the portable version you can drop into any future strategic conversation. It carries the same four voices but without the drill-specific phase references. The model name is a slot, so it adapts to whatever Claude model you run.

Save all rubric and persona files in your vault.
Commit all drill files to your core-drill branch and push to GitHub.
Merge the core-drill branch into main.
In Linear, close ticket 2 (Core drill: persona boot + step prompts) and ticket 3 (Rubrics and capstone).

Verify your vault now contains: drill/persona-boot.md, eight files in drill/steps/, drill/rubrics/per-step-rubric.md, drill/rubrics/capstone.md, and drill/persona-block-reusable.md.
Confirm both Linear tickets show as closed.

Capstone scorecard missing the harder variant pass bar?

The capstone must include both the default pass bar (8/12, items 3 and 5 scoring at least 1) and the harder variant pass bar (10/12, items 3 and 5 scoring 2). If Claude omitted the harder variant section, prompt it to add the 10/12 bar with the kill-decision defense requirement.

Help me check my capstone scorecard

✔️ Awesome, I've got everything!

Confirm your core-drill branch is merged and both tickets are closed. The drill engine is built.

ⓧ I'd like to double check the full capstone

Here is the complete capstone scorecard for comparison:

# Capstone Scorecard

## Six Items (scored 0-2 each, 12 maximum)

### 1. Genuine Where-to-Play Choice
- 0: No specific market, customer, or geography named. "We serve everyone."
- 1: A market is named but exclusions are vague or implicit.
- 2: Specific where-to-play with explicit exclusions. Names who is not served.

### 2. How-to-Win Advantage
- 0: Generic aspiration ("be the best," "deliver value"). Could apply to any competitor.
- 1: An advantage is named but not differentiated from competitors.
- 2: Names why you win and not a specific named competitor. The advantage is yours.

### 3. Real Trade-Off
- 0: Nothing is stopped or defunded. "We will do this and everything else."
- 1: Something is named as lower priority but not explicitly stopped or defunded.
- 2: Something specific is explicitly stopped or defunded. Named, with dollars or headcount freed.

### 4. Resources Actually Reallocated
- 0: No budget or headcount change from current state.
- 1: Resources are named but vague ("some people will shift").
- 2: Specific dollars and headcount moved, from named programs to named programs. The revealed-strategy test: if you read only the budget, you would know the strategy.

### 5. Human Cost and Trust Named
- 0: No mention of who the choice costs or what trust it requires.
- 1: Human cost is acknowledged in general terms but no specific person or role is named.
- 2: Names who is asked to change, who loses, what trust capital is spent, and whether it is safe to say the strategy is not working. Specific, not abstract.

### 6. Coherent Narrative
- 0: No narrative, or the narrative contradicts the allocation.
- 1: A narrative exists but a team would struggle to repeat it, or it partially contradicts the funded reality.
- 2: A two-sentence narrative a team can repeat, consistent with the funded reality. The Strategy Mirror confirms coherence.

## Pass Bar

Total: 8 of 12 minimum.
Required items: Item 3 (Real Trade-Off) and Item 5 (Human Cost and Trust Named) must each score at least 1.

A strategy that stops nothing fails regardless of total score.
A brief with a blank human-cost line fails regardless of total score.

## Harder Variant Pass Bar

Total: 10 of 12 minimum.
Required items: Items 3 and 5 must each score 2 (not just 1).
Additional: the kill decision must be defended (why this bet dies and not the other).

## Teach-Back (pass/fail)

The leader explains the bet, the trade-off, and the human cost to a skeptical peer in five minutes and defends what they stopped.
- Pass: the peer (or Strategy Mirror) can repeat the bet, the trade-off, and the cost accurately.
- Fail: the peer cannot repeat it, or the defense of what was stopped does not hold.

Your drill engine is built: four personas that refuse to let you skip the hard parts, eight phases timed to ninety minutes, and a scorecard that fails strategies without real trade-offs. Next, you will build the supporting artifacts that make the drill usable: worksheets, templates, variants, and validation instruments.

Build Supporting Artifacts and Validation Instruments

Your core drill system is complete. Four personas with refusal rules, eight timed phases, and a capstone scorecard are committed and merged. The engine that forces honest strategy is built.

Now you need the surfaces the drill writes on. Worksheets that structure the leader's thinking at each phase, a decision-brief template that collects the output, operational variants for different contexts, and a validation instrument that checks whether the bet survived contact with reality at 30 and 90 days.

In this step, get ready to:

Author three worksheets and the seven-field decision-brief template.
Build operational variants (lighter and harder) for different contexts.
Build the 30/90-day re-score instrument and the plain-language guide.

Author worksheets and the decision-brief template

Each phase of the drill needs a structured surface. The diagnosis phase needs prompts that prevent flattering reframes. The cascade phase needs the five Lafley/Martin questions in order. The reallocation phase needs tables that force specific numbers.

Have Claude generate the honest-diagnosis worksheet by prompting it to create worksheets/honest-diagnosis.md with the content below:

# Honest Diagnosis Worksheet

Source: Rumelt's kernel (Good Strategy Bad Strategy, 2011). A strategy begins with an honest diagnosis of the challenge, not a flattering reframe.

## The Real Problem

What is the real strategic problem? Not the polite version. Not the version you would put in a board deck. The one that keeps the current allocation from working.

[Answer here]

## What Has Failed

What has the current allocation failed to solve? Be specific. Name programs, initiatives, or investments that have not produced the result they promised.

[Answer here]

## The Uncomfortable Truth

What uncomfortable truth have you been avoiding? The thing that, if named, would require a different decision than the one currently funded.

[Answer here]

## The Outsider View

What would an honest outsider (someone with no political stake in your current allocation) say about your current strategy?

[Answer here]

## Diagnosis in One Sentence

The core challenge is: [one sentence that names the real problem without flattery]

---

Scoring note: The Strategist scores this phase. A polite evasion scores 0. A diagnosis that names the problem but hedges scores 1. A diagnosis that is specific, uncomfortable, and honest scores 2.

What does this worksheet do?

The honest-diagnosis worksheet operationalizes Rumelt's kernel. It forces the leader through five escalating prompts, each designed to strip away the polite framing that makes most strategy decks useless.

The final one-sentence diagnosis becomes the foundation for the entire cascade. If the diagnosis is flattering, every choice built on it will fail in execution.

Save the file to your vault at worksheets/honest-diagnosis.md.
Confirm the file renders in your markdown editor with all five sections visible.

Claude producing a generic worksheet?

If Claude generates vague prompts like "What is your strategy?" instead of the confrontational framing above, remind it that the worksheet must prevent flattering reframes. Specify in your prompt that each question should make politeness impossible.

help me get Claude to generate a worksheet that forces honest diagnosis instead of polite strategy language

Have Claude generate the where-to-play/how-to-win cascade worksheet at worksheets/cascade.md:

# Where-to-Play / How-to-Win Cascade Worksheet

Source: Lafley and Martin Playing-to-Win cascade (Playing to Win, 2013). Strategy is an integrated set of choices that positions you to win.

## Winning Aspiration

What does winning look like for this specific choice? Not a corporate mission statement. What does success look like in 12-18 months if this bet works?

[Answer here]

## Where to Play

Which customers, geographies, channels, segments, or categories?
Be specific. Name them.

Who is explicitly excluded? Who are you choosing NOT to serve with this bet?

[Answer here]

## How to Win

What is the advantage? Why do you win here and not a competitor?
Name the competitor. Name why you beat them in this specific where-to-play.

If this advantage could apply to any company in your industry, it is not an advantage. Start over.

[Answer here]

## Capabilities Required

What must be in place to win here? What capabilities do you currently have? What do you lack and must build or acquire?

[Answer here]

## Management Systems

What changes in reporting, incentives, meetings, governance, or operating rhythm to support this bet? What current system actively works against it?

[Answer here]

---

Scoring note: The Strategist applies the refusal rule here. If where-to-play names no exclusion, or how-to-win is generic, the drill halts until fixed.

What does this worksheet do?

The cascade worksheet walks the leader through all five choices in the Lafley/Martin framework in sequence. Each section builds on the previous one.

The critical guardrail is in the "How to Win" section: if the stated advantage could apply to any company in the industry, the leader must start over. This is where the Strategist persona's refusal rule fires.

Save the file to your vault at worksheets/cascade.md.
Confirm all five cascade sections are present and in the correct order (aspiration, where, how, capabilities, systems).

Have Claude generate the resource-reallocation worksheet at worksheets/resource-reallocation.md:

# Resource Reallocation Worksheet

Source: Resource allocation as revealed strategy. Porter's discipline: strategy is choosing what not to do (What Is Strategy?, HBR, 1996). If the money does not move, nothing has been decided.

## What We Fund

| Program / Initiative | Dollar Amount | Headcount | Timeline | Owner |
|---------------------|--------------|-----------|----------|-------|
| | | | | |
| | | | | |
| | | | | |

## What We Stop

| Program / Initiative | Dollars Freed | Headcount Freed | Who Owns the Stop | Why This One |
|---------------------|--------------|-----------------|-------------------|-------------|
| | | | | |
| | | | | |
| | | | | |

## Net Reallocation

Total dollars moved: [amount]
Total headcount shifted: [number]
From: [named programs]
To: [named programs]

## The Revealed-Strategy Test

If someone read only our budget and headcount plan (with no narrative, no deck, no explanation), would they know our strategy? Would they correctly guess our where-to-play and how-to-win from the allocation alone?

[Answer: yes/no and why]

## What Will Creep Back

Name the thing we said we would stop that will try to return. Who will champion its return? What is our answer when they do?

[Answer here]

---

Scoring note: The Board Chair applies the refusal rule here. If nothing real moves (no dollars, no headcount, no programs stopped), the drill halts. "Efficiency gains" and "we will find the budget" are not reallocation.

Why tables instead of open prompts?

The reallocation worksheet uses tables with specific columns (dollar amounts, headcount, owners) because vague reallocation is the most common failure mode. A leader who writes "we will shift resources" has not made a decision. A leader who writes "$2.4M from Program X, 6 headcount from Team Y" has.

The "What Will Creep Back" section is unique to this worksheet. It forces the leader to name the stopped program that will try to resurrect itself. Naming it now makes it harder to quietly restart later.

Save the file to your vault at worksheets/resource-reallocation.md.
Confirm both tables render correctly in your markdown editor with all column headers visible.

Have Claude generate the one-page decision-brief template at brief/template.md:

# One-Page Strategic Decision Brief

## 1. The Bet
[One sentence. What we are choosing to do and where we are playing.]

## 2. What Is Funded
[Specific programs, dollar amounts, headcount allocated. Name them.]

## 3. What Is Stopped
[Specific programs defunded or ended. Dollar amounts and headcount freed. Who owns the stop.]

## 4. Who Leads It
[Named person. Their authority. Their accountability. What they can decide without escalating.]

## 5. Human Cost and Trust Required
[Who is asked to change. Who loses. What trust capital is being spent. Is it safe to say this is not working? How do they tell you?]

## 6. The Narrative
[Two sentences maximum. The story a team can repeat. Must match the funded reality.]

## 7. The Metric
[How we know by day 90 whether the bet is working. One metric, measurable, with a threshold.]

---

## Capstone Score: __ / 12
- Item 3 (Trade-Off): __ / 2
- Item 5 (Human Cost): __ / 2
- Pass: Yes / No

## Teach-Back: Pass / Fail

## Cadence Booked:
- Weekly check-in: [day/time]
- 30-day re-score: [date]
- 90-day re-score: [date]

What does this template do?

This is the signature artifact of the drill. All eight phases converge into these seven fields. The Brief Compiler agent in Phase 8 aggregates your worksheet answers into this single page.

The capstone score, teach-back result, and cadence fields at the bottom make the brief self-validating. A brief that passes the capstone and the teach-back is a brief a team can execute.

Save the file to your vault at brief/template.md.
Confirm all seven numbered fields are present, plus the capstone score section and cadence section at the bottom.

Template missing fields?

If Claude produces fewer than seven fields, check that your prompt explicitly names all seven: the bet, what is funded, what is stopped, who leads it, human cost and trust, the narrative, and the metric. The most commonly dropped fields are "Who Leads It" and "The Narrative" because Claude sometimes folds them into other fields.

help me get Claude to produce a complete seven-field decision brief template

✔️ Awesome, I've got everything!

Confirm all four files are saved in your vault: three in worksheets/ and one in brief/.

ⓧ I'd like to double check the full code

Here are all four files. Compare your vault contents against these:

# Honest Diagnosis Worksheet

Source: Rumelt's kernel (Good Strategy Bad Strategy, 2011). A strategy begins with an honest diagnosis of the challenge, not a flattering reframe.

## The Real Problem

What is the real strategic problem? Not the polite version. Not the version you would put in a board deck. The one that keeps the current allocation from working.

[Answer here]

## What Has Failed

What has the current allocation failed to solve? Be specific. Name programs, initiatives, or investments that have not produced the result they promised.

[Answer here]

## The Uncomfortable Truth

What uncomfortable truth have you been avoiding? The thing that, if named, would require a different decision than the one currently funded.

[Answer here]

## The Outsider View

What would an honest outsider (someone with no political stake in your current allocation) say about your current strategy?

[Answer here]

## Diagnosis in One Sentence

The core challenge is: [one sentence that names the real problem without flattery]

---

Scoring note: The Strategist scores this phase. A polite evasion scores 0. A diagnosis that names the problem but hedges scores 1. A diagnosis that is specific, uncomfortable, and honest scores 2.

# Where-to-Play / How-to-Win Cascade Worksheet

Source: Lafley and Martin Playing-to-Win cascade (Playing to Win, 2013). Strategy is an integrated set of choices that positions you to win.

## Winning Aspiration

What does winning look like for this specific choice? Not a corporate mission statement. What does success look like in 12-18 months if this bet works?

[Answer here]

## Where to Play

Which customers, geographies, channels, segments, or categories?
Be specific. Name them.

Who is explicitly excluded? Who are you choosing NOT to serve with this bet?

[Answer here]

## How to Win

What is the advantage? Why do you win here and not a competitor?
Name the competitor. Name why you beat them in this specific where-to-play.

If this advantage could apply to any company in your industry, it is not an advantage. Start over.

[Answer here]

## Capabilities Required

What must be in place to win here? What capabilities do you currently have? What do you lack and must build or acquire?

[Answer here]

## Management Systems

What changes in reporting, incentives, meetings, governance, or operating rhythm to support this bet? What current system actively works against it?

[Answer here]

---

Scoring note: The Strategist applies the refusal rule here. If where-to-play names no exclusion, or how-to-win is generic, the drill halts until fixed.

# Resource Reallocation Worksheet

Source: Resource allocation as revealed strategy. Porter's discipline: strategy is choosing what not to do (What Is Strategy?, HBR, 1996). If the money does not move, nothing has been decided.

## What We Fund

| Program / Initiative | Dollar Amount | Headcount | Timeline | Owner |
|---------------------|--------------|-----------|----------|-------|
| | | | | |
| | | | | |
| | | | | |

## What We Stop

| Program / Initiative | Dollars Freed | Headcount Freed | Who Owns the Stop | Why This One |
|---------------------|--------------|-----------------|-------------------|-------------|
| | | | | |
| | | | | |
| | | | | |

## Net Reallocation

Total dollars moved: [amount]
Total headcount shifted: [number]
From: [named programs]
To: [named programs]

## The Revealed-Strategy Test

If someone read only our budget and headcount plan (with no narrative, no deck, no explanation), would they know our strategy? Would they correctly guess our where-to-play and how-to-win from the allocation alone?

[Answer: yes/no and why]

## What Will Creep Back

Name the thing we said we would stop that will try to return. Who will champion its return? What is our answer when they do?

[Answer here]

---

Scoring note: The Board Chair applies the refusal rule here. If nothing real moves (no dollars, no headcount, no programs stopped), the drill halts. "Efficiency gains" and "we will find the budget" are not reallocation.

# One-Page Strategic Decision Brief

## 1. The Bet
[One sentence. What we are choosing to do and where we are playing.]

## 2. What Is Funded
[Specific programs, dollar amounts, headcount allocated. Name them.]

## 3. What Is Stopped
[Specific programs defunded or ended. Dollar amounts and headcount freed. Who owns the stop.]

## 4. Who Leads It
[Named person. Their authority. Their accountability. What they can decide without escalating.]

## 5. Human Cost and Trust Required
[Who is asked to change. Who loses. What trust capital is being spent. Is it safe to say this is not working? How do they tell you?]

## 6. The Narrative
[Two sentences maximum. The story a team can repeat. Must match the funded reality.]

## 7. The Metric
[How we know by day 90 whether the bet is working. One metric, measurable, with a threshold.]

---

## Capstone Score: __ / 12
- Item 3 (Trade-Off): __ / 2
- Item 5 (Human Cost): __ / 2
- Pass: Yes / No

## Teach-Back: Pass / Fail

## Cadence Booked:
- Weekly check-in: [day/time]
- 30-day re-score: [date]
- 90-day re-score: [date]

Build operational variants and the reusable persona block

The default drill runs 90 minutes on one bet. Different contexts need different intensities. A lighter variant for warm-ups or lower-stakes decisions. A harder variant that forces the kill decision between competing bets.

What about the reusable persona block?

The reusable persona block at drill/persona-block-reusable.md was extracted in the previous step when you built the core drill. It is already in your vault and committed. No action needed here.

Have Claude generate the lighter variant at variants/lighter.md:

# Lighter Variant (60 min)

60-minute version on a smaller unit-level choice. Skips Phase 7 (pressure test). Cannot satisfy the capstone (which requires the full challenge). Use as a warm-up or for lower-stakes decisions.

Why can the lighter variant not satisfy the capstone?

The capstone scorecard requires the pressure test (Phase 7) to validate coherence under challenge. Without it, the leader has not defended what they stopped or proved the narrative holds under scrutiny. The lighter variant is useful for practice and lower-stakes choices, but it cannot produce a passing score.

Save the file to your vault at variants/lighter.md.

Have Claude generate the harder variant at variants/harder.md:

# Harder Variant: Competing Bets

## What This Is

The default drill forces one bet. The harder variant forces two competing bets on the same strategic problem, then makes the leader kill one on the evidence. It tests whether you can make a trade-off between two things you want, not just between something good and something dead.

## How It Runs

Phases 1-6 run twice, forked at the cascade phase:
- Bet A gets a full cascade, full reallocation, full human-cost analysis.
- Bet B gets the same.
- Both are defensible. Both have resources behind them. Both cost someone.

At Phase 7 (Pressure Test), the Board Chair forces the kill:
"You cannot fund both. Which one dies? Show me why."

The leader must:
1. Present the evidence for both bets (what each would win, what each costs).
2. Choose one to kill. Explicitly. Not "deprioritize." Kill.
3. Name what you lose by killing it.
4. Name who was championing it and what this costs them (the human cost of the kill).
5. Defend the surviving bet against the Board Chair: "Why this one and not the other?"

## Pass Bar (Harder)

Total: 10 of 12 minimum.
Items 3 (real trade-off) and 5 (human cost) must each score 2 (not just 1).
Additional: the kill decision must be defended. "Why does this bet die and not the other?" must have a specific, evidence-based answer.

## What It Reveals

- If you cannot name a second credible bet on the same problem, your where-to-play choice may not have been a real choice. There was only one plausible path, which means you diagnosed, you did not choose.
- If killing one feels costless, neither bet had real commitment behind it.
- The kill should hurt. That is the point.
- A leader who cannot kill their own good idea has not yet made a strategic choice.

## Scoring the Kill

The kill decision scores on:
- Specificity: is the killed bet named, specific, and defensible (not a straw man)?
- Evidence: is the decision based on evidence, not politics or fatigue?
- Human cost: is the person who championed the killed bet named, and is the cost to them acknowledged?
- Finality: is it dead, or is it "paused" (which means alive)?

What does the harder variant test?

The default drill tests whether you can make a bet. The harder variant tests whether you can kill your own good idea. Most leaders fail here because they "deprioritize" instead of stopping. The variant's language is deliberate: kill, not pause.

The higher pass bar (10/12 with items 3 and 5 both scoring 2) means the leader cannot hedge on trade-offs or human cost. Partial honesty does not pass the harder variant.

Save the file to your vault at variants/harder.md.
Confirm both variant files are in the variants/ folder.

Harder variant missing the kill mechanics?

If Claude produces a harder variant that simply "adds more phases" or "increases difficulty" without the competing-bets-then-kill structure, your prompt needs to specify: two full bets through the cascade, then a forced kill at Phase 7 with the Board Chair demanding the decision. The kill is the whole point of the harder variant.

help me prompt Claude to produce a competing-bets variant that forces the kill decision

Build validation instruments and the plain-language guide

A decision brief without follow-up is a one-time exercise. The re-score instrument turns a single drill into a leadership practice by checking whether the money actually moved at 30 and 90 days.

Have Claude generate the 30/90-day re-score instrument at drill/rubrics/re-score.md:

# 30 and 90-Day Re-Score Instrument

## Weekly Check-In (5 minutes)

Three questions, every week:
1. Did the money and people actually move this week? What specifically shifted?
2. What did we stop that is creeping back? Who is championing its return?
3. Is the team still repeating the narrative? (Ask someone. Do not assume.)

## Day 30 Re-Score

Score the same six capstone items against reality (not intent):

1. Where-to-play choice: Are we actually playing there? Or have we drifted back to "everyone"?
2. How-to-win advantage: Is the advantage being built? Or are we spreading across everything again?
3. Real trade-off: Is the stopped thing still stopped? Has anything been quietly restarted?
4. Resources reallocated: Did the money actually move in the financial system? Did the headcount actually shift? Check the actuals, not the plan.
5. Human cost and trust: Did the people whose Monday changed actually change? Do they trust the leader? Is it safe to say it is not working? (Ask them, do not assume.)
6. Coherent narrative: Can a team member two levels down repeat the bet and the trade-off accurately?

Pass bar (day 30): Same as capstone. 8/12, items 3 and 5 scoring. If the allocation has reverted, the strategy has failed regardless of what the deck says.

## Day 90 Re-Score

Everything from day 30, plus:
- Does the metric show a first result? What is the number?
- If the metric has not moved, does the evidence warrant re-running the drill on a revised diagnosis?
- Has the decision brief become how this leader makes strategic calls, or was it a one-time exercise?

Decision point: If day 90 shows reversion (money back where it was, stopped programs restarted, team cannot repeat the narrative), the leader re-runs the drill on the evidence. The question is not "was the bet right?" but "was it a bet at all?"

## Day 180 Check (informal)

- Has the money stayed moved?
- Is the team still repeating why?
- Has the leader stopped something else using the same method?
- The real measure is not one clean deck but money and people that actually moved, a team that can repeat why, and the trust that comes from a leader who stopped something and said who it cost.

Why score against reality, not intent?

The re-score instrument asks "did the money actually move in the financial system?" not "do you plan to move it?" This is the revealed-strategy test applied over time. Most strategies revert within 30 days because the stopped program quietly restarts.

The day 90 decision point is binary: if the allocation has reverted, the strategy has failed. It does not matter what the deck says or what the leader intended. Where the money sits is the real strategy.

Save the file to your vault at drill/rubrics/re-score.md.
Confirm the file contains all four time horizons: weekly, day 30, day 90, and day 180.

Have Claude generate the plain-language guide at guide/guide.md:

# Set the Bet: Wharton Strategy, 90 Minutes

## Plain-Language Guide

## What This Is

A 90-minute AI-facilitated leadership drill that produces a one-page strategic decision brief on a real choice you own. Four AI personas coach, fund, and challenge. The output is not a deck. It is a decision a team can execute.

This is the operator's design estimate that a ninety-minute exercise transfers much of the operational core of a roughly $79,000 executive program's strategic-leadership curriculum (programs like those at Wharton, INSEAD, and London Business School). It is not a guarantee. The credential, the community, and the supervised practice are what those programs still uniquely provide.

## Why It Exists

Strategic leadership becomes real only when applied to a live decision the leader owns and can move money and people behind. Most strategy processes produce lists of priorities that fund nothing differently and stop nothing. This drill forces the bet.

## What You Need

- A flat-rate Claude client (Claude Code or Cowork). No metered API. The model name is a slot; use whatever current Claude model you run.
- A plain markdown editor (Obsidian recommended) for the brief and worksheets.
- 90 uninterrupted minutes.
- One real strategic choice you own this quarter, where you can actually move money and people.
- The honesty to name what it costs and who it costs.

## How to Run It in Ninety Minutes

| Phase | Time | What Happens |
|-------|------|-------------|
| 1. Setup and the Real Choice | 5 min | Name the one real bet. Confirm it is yours, it is real, and you can move resources. |
| 2. Boot the Cast | 5 min | Four personas activate. Each states its refusal rule and asks one opening question. |
| 3. Honest Diagnosis | 10 min | Name the real problem. Fill the diagnosis worksheet. No flattering reframes. |
| 4. Where-to-Play / How-to-Win | 15 min | Run the cascade. Fill the worksheet. The Strategist halts if the choice is not specific. |
| 5. Fund the Bet | 15 min | Reallocate money and people. Fill the reallocation worksheet. The Board Chair halts if nothing moves. |
| 6. Lead the Change | 15 min | Name the human cost, the trust required, the safety to dissent. The Frontline Voice halts if the line is blank. |
| 7. Pressure Test | 10 min | Board Chair and Frontline Voice challenge together. Strategy Mirror checks coherence. |
| 8. Decision Brief and Capstone | 15 min | Compile the brief. Score the capstone. Run the teach-back. Book the cadence. |

## How the Personas Work

Each persona has a refusal rule: a condition under which it halts the drill and forces you back. This is by design. A rubric at the end lets bad moves through. Refusal rules stop you mid-phase when you try to skip the hard part.

- The Strategist halts when you have not made a specific choice (who is excluded, what you are not doing).
- The Board Chair halts when nothing real moves (no dollars, no headcount, no programs stopped).
- The Frontline Voice halts when the human-cost line is blank (no person named, no trust calculus, no safety).
- The Strategy Mirror flags (does not halt) when the narrative contradicts the allocation.

## How the Rubrics Work

Per-step: each phase scores 0 (evasive), 1 (partial), or 2 (specific and honest). Refusal rules override: if triggered, phase scores 0 until fixed.

Capstone: six items, 0-2 each, 12 maximum. Pass at 8/12. Items 3 (real trade-off) and 5 (human cost) must both score at least 1. A strategy that stops nothing or ignores who it costs fails.

## Variants

- Lighter (60 min): smaller unit-level choice, skips pressure test, cannot satisfy capstone.
- Default (90 min): full drill on a real enterprise or division choice.
- Harder: two competing bets, force the kill, higher pass bar (10/12).

## Where the Deeper Material Lives

The recognized canon this drill draws on:
- Rumelt, Richard. Good Strategy Bad Strategy (2011). The kernel: diagnosis, guiding policy, coherent action.
- Lafley, A.G. and Martin, Roger. Playing to Win (2013). The cascade: winning aspiration, where to play, how to win, capabilities, management systems.
- Porter, Michael. "What Is Strategy?" Harvard Business Review (1996). Strategy is choosing what not to do.
- Resource allocation as revealed strategy: where the money goes is the real strategy.
- Teece, David. "Dynamic Capabilities and Strategic Management" (1997). Sense, seize, transform.

## The Value Claim

This is the operator's design estimate that a ninety-minute exercise transfers much of the operational core of a roughly $79,000 executive program's strategic-leadership curriculum. It is not a guarantee. The credential, the community, and the supervised practice are what the program itself still uniquely provides.

The drill teaches judgment, not a right answer. Two leaders will make different defensible bets. It names the human cost but cannot remove it.

What does the guide cover?

The guide is the entry point for anyone running the drill for the first time or revisiting it later. It covers what the drill is, why it exists, what you need, how to run it in ninety minutes, how personas and rubrics work, the three variants, the source canon, and the value claim.

The phase timing table gives a quick-reference overview. The persona summary shows each refusal rule in plain language. Someone reading only the guide should understand the full system without opening any other file.

Save the file to your vault at guide/guide.md.
Confirm the guide contains all required sections: What This Is, Why It Exists, What You Need, How to Run It, How the Personas Work, How the Rubrics Work, Variants, Where the Deeper Material Lives, and The Value Claim.

Guide missing the value claim or source framing?

These sections are deliberately included. The value claim sets honest expectations (design estimate, not a guarantee). The source framing grounds the drill in the recognized canon so the reader knows this is not invented methodology. If Claude omits them, explicitly request both sections with the verbatim framing from principles.md.

help me get Claude to include the value claim and source canon in the guide

Commit all new files to your GitHub repo on a supporting-artifacts branch.
Merge the branch to main.
Return to your Linear project and close tickets 4 (Worksheets and brief template), 5 (Variants and reusable persona block), 6 (Validation instruments), and 7 (Plain-language guide).

✔️ Awesome, I've got everything!

Confirm your branch is merged and all four Linear tickets are closed.

ⓧ I'd like to double check the full code

Your vault should now contain these new files (in addition to everything from previous steps):

worksheets/honest-diagnosis.md
worksheets/cascade.md
worksheets/resource-reallocation.md
brief/template.md
variants/lighter.md
variants/harder.md
drill/rubrics/re-score.md
guide/guide.md

Compare each file against the code blocks above. The content should match exactly.

# Lighter Variant (60 min)

60-minute version on a smaller unit-level choice. Skips Phase 7 (pressure test). Cannot satisfy the capstone (which requires the full challenge). Use as a warm-up or for lower-stakes decisions.

# Harder Variant: Competing Bets

## What This Is

The default drill forces one bet. The harder variant forces two competing bets on the same strategic problem, then makes the leader kill one on the evidence. It tests whether you can make a trade-off between two things you want, not just between something good and something dead.

## How It Runs

Phases 1-6 run twice, forked at the cascade phase:
- Bet A gets a full cascade, full reallocation, full human-cost analysis.
- Bet B gets the same.
- Both are defensible. Both have resources behind them. Both cost someone.

At Phase 7 (Pressure Test), the Board Chair forces the kill:
"You cannot fund both. Which one dies? Show me why."

The leader must:
1. Present the evidence for both bets (what each would win, what each costs).
2. Choose one to kill. Explicitly. Not "deprioritize." Kill.
3. Name what you lose by killing it.
4. Name who was championing it and what this costs them (the human cost of the kill).
5. Defend the surviving bet against the Board Chair: "Why this one and not the other?"

## Pass Bar (Harder)

Total: 10 of 12 minimum.
Items 3 (real trade-off) and 5 (human cost) must each score 2 (not just 1).
Additional: the kill decision must be defended. "Why does this bet die and not the other?" must have a specific, evidence-based answer.

## What It Reveals

- If you cannot name a second credible bet on the same problem, your where-to-play choice may not have been a real choice. There was only one plausible path, which means you diagnosed, you did not choose.
- If killing one feels costless, neither bet had real commitment behind it.
- The kill should hurt. That is the point.
- A leader who cannot kill their own good idea has not yet made a strategic choice.

## Scoring the Kill

The kill decision scores on:
- Specificity: is the killed bet named, specific, and defensible (not a straw man)?
- Evidence: is the decision based on evidence, not politics or fatigue?
- Human cost: is the person who championed the killed bet named, and is the cost to them acknowledged?
- Finality: is it dead, or is it "paused" (which means alive)?

# 30 and 90-Day Re-Score Instrument

## Weekly Check-In (5 minutes)

Three questions, every week:
1. Did the money and people actually move this week? What specifically shifted?
2. What did we stop that is creeping back? Who is championing its return?
3. Is the team still repeating the narrative? (Ask someone. Do not assume.)

## Day 30 Re-Score

Score the same six capstone items against reality (not intent):

1. Where-to-play choice: Are we actually playing there? Or have we drifted back to "everyone"?
2. How-to-win advantage: Is the advantage being built? Or are we spreading across everything again?
3. Real trade-off: Is the stopped thing still stopped? Has anything been quietly restarted?
4. Resources reallocated: Did the money actually move in the financial system? Did the headcount actually shift? Check the actuals, not the plan.
5. Human cost and trust: Did the people whose Monday changed actually change? Do they trust the leader? Is it safe to say it is not working? (Ask them, do not assume.)
6. Coherent narrative: Can a team member two levels down repeat the bet and the trade-off accurately?

Pass bar (day 30): Same as capstone. 8/12, items 3 and 5 scoring. If the allocation has reverted, the strategy has failed regardless of what the deck says.

## Day 90 Re-Score

Everything from day 30, plus:
- Does the metric show a first result? What is the number?
- If the metric has not moved, does the evidence warrant re-running the drill on a revised diagnosis?
- Has the decision brief become how this leader makes strategic calls, or was it a one-time exercise?

Decision point: If day 90 shows reversion (money back where it was, stopped programs restarted, team cannot repeat the narrative), the leader re-runs the drill on the evidence. The question is not "was the bet right?" but "was it a bet at all?"

## Day 180 Check (informal)

- Has the money stayed moved?
- Is the team still repeating why?
- Has the leader stopped something else using the same method?
- The real measure is not one clean deck but money and people that actually moved, a team that can repeat why, and the trust that comes from a leader who stopped something and said who it cost.

# Set the Bet: Wharton Strategy, 90 Minutes

## Plain-Language Guide

## What This Is

A 90-minute AI-facilitated leadership drill that produces a one-page strategic decision brief on a real choice you own. Four AI personas coach, fund, and challenge. The output is not a deck. It is a decision a team can execute.

This is the operator's design estimate that a ninety-minute exercise transfers much of the operational core of a roughly $79,000 executive program's strategic-leadership curriculum (programs like those at Wharton, INSEAD, and London Business School). It is not a guarantee. The credential, the community, and the supervised practice are what those programs still uniquely provide.

## Why It Exists

Strategic leadership becomes real only when applied to a live decision the leader owns and can move money and people behind. Most strategy processes produce lists of priorities that fund nothing differently and stop nothing. This drill forces the bet.

## What You Need

- A flat-rate Claude client (Claude Code or Cowork). No metered API. The model name is a slot; use whatever current Claude model you run.
- A plain markdown editor (Obsidian recommended) for the brief and worksheets.
- 90 uninterrupted minutes.
- One real strategic choice you own this quarter, where you can actually move money and people.
- The honesty to name what it costs and who it costs.

## How to Run It in Ninety Minutes

| Phase | Time | What Happens |
|-------|------|-------------|
| 1. Setup and the Real Choice | 5 min | Name the one real bet. Confirm it is yours, it is real, and you can move resources. |
| 2. Boot the Cast | 5 min | Four personas activate. Each states its refusal rule and asks one opening question. |
| 3. Honest Diagnosis | 10 min | Name the real problem. Fill the diagnosis worksheet. No flattering reframes. |
| 4. Where-to-Play / How-to-Win | 15 min | Run the cascade. Fill the worksheet. The Strategist halts if the choice is not specific. |
| 5. Fund the Bet | 15 min | Reallocate money and people. Fill the reallocation worksheet. The Board Chair halts if nothing moves. |
| 6. Lead the Change | 15 min | Name the human cost, the trust required, the safety to dissent. The Frontline Voice halts if the line is blank. |
| 7. Pressure Test | 10 min | Board Chair and Frontline Voice challenge together. Strategy Mirror checks coherence. |
| 8. Decision Brief and Capstone | 15 min | Compile the brief. Score the capstone. Run the teach-back. Book the cadence. |

## How the Personas Work

Each persona has a refusal rule: a condition under which it halts the drill and forces you back. This is by design. A rubric at the end lets bad moves through. Refusal rules stop you mid-phase when you try to skip the hard part.

- The Strategist halts when you have not made a specific choice (who is excluded, what you are not doing).
- The Board Chair halts when nothing real moves (no dollars, no headcount, no programs stopped).
- The Frontline Voice halts when the human-cost line is blank (no person named, no trust calculus, no safety).
- The Strategy Mirror flags (does not halt) when the narrative contradicts the allocation.

## How the Rubrics Work

Per-step: each phase scores 0 (evasive), 1 (partial), or 2 (specific and honest). Refusal rules override: if triggered, phase scores 0 until fixed.

Capstone: six items, 0-2 each, 12 maximum. Pass at 8/12. Items 3 (real trade-off) and 5 (human cost) must both score at least 1. A strategy that stops nothing or ignores who it costs fails.

## Variants

- Lighter (60 min): smaller unit-level choice, skips pressure test, cannot satisfy capstone.
- Default (90 min): full drill on a real enterprise or division choice.
- Harder: two competing bets, force the kill, higher pass bar (10/12).

## Where the Deeper Material Lives

The recognized canon this drill draws on:
- Rumelt, Richard. Good Strategy Bad Strategy (2011). The kernel: diagnosis, guiding policy, coherent action.
- Lafley, A.G. and Martin, Roger. Playing to Win (2013). The cascade: winning aspiration, where to play, how to win, capabilities, management systems.
- Porter, Michael. "What Is Strategy?" Harvard Business Review (1996). Strategy is choosing what not to do.
- Resource allocation as revealed strategy: where the money goes is the real strategy.
- Teece, David. "Dynamic Capabilities and Strategic Management" (1997). Sense, seize, transform.

## The Value Claim

This is the operator's design estimate that a ninety-minute exercise transfers much of the operational core of a roughly $79,000 executive program's strategic-leadership curriculum. It is not a guarantee. The credential, the community, and the supervised practice are what the program itself still uniquely provides.

The drill teaches judgment, not a right answer. Two leaders will make different defensible bets. It names the human cost but cannot remove it.

Your drill system now has everything it needs: worksheets to structure the thinking, a brief template to collect the output, variants for different contexts, a re-score instrument to check whether the bet survived, and a guide that lets anyone run it. Next up, you run the drill on the real choice you own.

Run the Drill and Ship the Release

Steps 1 through 4 built the complete system: four personas with refusal rules, eight timed phases, rubrics, worksheets, and a capstone that fails any strategy without a real trade-off or a named human cost. The architecture is committed, the guide is written, and the tickets are closing.

Now you use the system on the real choice you named at the start. This is where leadership happens. The AI facilitated the build. You lead the decision.

In this step, get ready to:

Execute the 90-minute drill on your real strategic choice, filling all three worksheets.
Score the capstone, run the teach-back, and book the cadence.
Deliver the readout, document lessons learned, and tag the release.

Execute the 90-minute drill on your real choice

The drill is not a thought experiment. It is a leadership act on a real bet you own this quarter, where you can actually move money and people.

Return to your Claude client.
Load the full content of drill/persona-boot.md into the conversation. Paste the entire file so Claude can boot all four personas.
Tell Claude to begin Phase 1: Setup and the Real Choice.

How the 90 minutes break down

Run all eight phases in sequence. Respect the timing. When a persona fires a refusal rule, stop and fix the gap before advancing.

Phase 1 (5 min): Name your one real strategic choice. Confirm it is yours, it is real, and you can move resources.
Phase 2 (5 min): Boot the four personas. Each states its refusal rule and asks one opening question.
Phase 3 (10 min): Honest diagnosis. Fill worksheets/honest-diagnosis.md as you go.
Phase 4 (15 min): Where-to-play and how-to-win cascade. Fill worksheets/cascade.md as you go.
Phase 5 (15 min): Fund the bet. Fill worksheets/resource-reallocation.md with real dollars, headcount, and programs stopped.
Phase 6 (15 min): Lead the change. Name who is asked to change, who loses, what trust capital you are spending, and whether it is safe for them to say it is not working.
Phase 7 (10 min): Pressure test. The Board Chair and the Frontline Voice challenge together. The Strategy Mirror checks coherence.
Phase 8 (15 min): Compile the decision brief, score the capstone, run the teach-back, book the cadence.

As you complete each phase, ask Claude to score it using the per-step rubric (0, 1, or 2).
If a persona fires its refusal rule, the phase scores 0. Go back and fix the gap before moving forward.
Save each worksheet in your vault as you complete it.

If a refusal rule fires

This is the drill working, not a failure. The Strategist halts when you have not excluded anyone. The Board Chair halts when nothing real moves. The Frontline Voice halts when the human-cost line is blank. Go back to the gap and name what you were avoiding.

At the end of Phase 8, ask Claude to compile your answers into the seven-field decision brief.
Create brief/completed-brief.md in your vault using the same structure as brief/template.md, with all seven fields filled from your real choice.
Confirm all seven fields are filled: The Bet, What Is Funded, What Is Stopped, Who Leads It, Human Cost and Trust Required, The Narrative, and The Metric.

Stuck on a phase?

If you cannot fill a field, that is diagnostic information. The drill is surfacing a gap in your strategy. Do not write a placeholder. Either name the real answer (uncomfortable as it may be) or acknowledge that you do not yet have one and note what you need to resolve it.

Ask Claude: I am stuck on [field]. Help me identify what I am avoiding naming.

Validate and teach back

The capstone scorecard is the test of whether your brief is a strategy or a list. Two items carry the pass bar: the real trade-off and the human cost.

Ask Claude to score your completed brief against the six capstone items (0-2 each, 12 maximum).
Confirm the pass bar: total of 8/12 or above, with Item 3 (Real Trade-Off) and Item 5 (Human Cost and Trust Named) both scoring at least 1.

What if you fail the capstone?

If your total is below 8, or if Item 3 or Item 5 scores 0, the drill has done its job. It showed you where your strategy is still a list. Go back to the phase where the refusal rule would have fired and redo it honestly.

A strategy that stops nothing fails regardless of total score. A brief with a blank human-cost line fails regardless of total score.

Run the teach-back: explain the bet, the trade-off, and the human cost to the Strategy Mirror persona (or a real peer) in five minutes. Defend what you stopped.
The teach-back passes if the listener can repeat the bet, the trade-off, and the cost accurately.

Book the cadence in your calendar app:
Create a recurring weekly check-in (5 minutes): Did the money and people actually move? What did we stop that is creeping back? Is the team still repeating the narrative?
Book the 30-day re-score (specific date, 30 days from today).
Book the 90-day re-score (specific date, 90 days from today).

Why book the cadence now?

The cadence is the test of whether the strategy survives contact with reality. If the money has not moved by day 30, the strategy has failed regardless of what the deck says. The 90-day re-score checks whether the metric shows a first result.

Fill in the Cadence Booked section at the bottom of brief/completed-brief.md with the actual days and dates you booked.
Save the file.

Close the build

The decision is made. Now close the build: deliver the readout, document what you learned, tag the release, and close the tickets.

Ask Claude to draft the two-audience readout and save it to readout/presentation.md. Two audiences, two framings:

Audience 1 (strategic peers): the bet, the cascade logic, the resources moved, the trade-off, and why this is a strategy and not a list.
Audience 2 (the team affected): the narrative, the human cost named honestly, the trust being asked for, and what safety looks like (how they tell you it is not working).

Ask Claude to draft readout/lessons-learned.md documenting four things:

What the drill surfaced that was not visible before.
Where refusal rules fired and what that revealed.
What you avoided naming until forced.
What changes in how you make strategic calls going forward.

Review both files in your editor and confirm they are honest, not performative.

Commit the drill run artifacts to your GitHub repo by running these commands:

git add .
git commit -m "Complete drill run: brief, readout, lessons learned"

Tag the release and push by running:

git tag -a v1.0.0 -m "Set the Bet v1.0: complete drill system and first run"
git push origin main
git push origin v1.0.0

What does the tag do?

The tag marks a specific commit as a release. Anyone (including your future self) can check out v1.0.0 to get the complete drill system as it stood when you first ran it. The repo README covers what this is, how to run it, what you need, and where things live.

Confirm the tag exists by running:

git tag

You should see v1.0.0 in the output.

Switch to your Linear project.
Close ticket 8 (Drill run and completed brief).
Close ticket 9 (Readout and lessons learned).
Close ticket 10 (Tagged release and repo docs).
Confirm all 10 tickets are closed.

Push rejected or tag already exists?

If you see a rejection on push, confirm you committed all changes first. If the tag name already exists from a test run, delete the old tag locally and remotely before re-creating it.

Ask Claude: My git push or tag command failed. Help me resolve it.

Your proof-of-work is complete: a decision brief with all seven fields filled, a capstone score that passes, a teach-back that holds, a booked cadence that will check reality, and a tagged release anyone can run.

Secret mission

The Harder Variant: Competing Bets

The default drill forced one bet. The harder variant forces two competing bets on the same strategic problem through the full cascade, then makes you kill one on the evidence. It tests whether you can make a trade-off between two things you want, not just between something good and something obviously dead.

Clean Up Your Resources

This project runs entirely on free-tier services and local files. There are no ongoing costs or charges to worry about.

Decide whether to keep your resources as permanent leadership infrastructure, pause the cadence until your next strategic bet, or archive everything.

Resources you used:

Local set-the-bet/ vault folder (all markdown files, worksheets, completed brief).
GitHub private repo (tagged release with all branches merged).
Linear project ("Set the Bet" with all tickets closed).
Calendar entries (weekly check-in, 30-day re-score, 90-day re-score).

Keep everything running

No action needed. This system is designed to be permanent infrastructure for how you make strategic calls from now on.

Keep the GitHub repo as your reusable drill system. Load the persona boot prompt for future strategic decisions.
Keep the Linear project for tracking future drill runs or new bets.
Keep the calendar cadence active. The weekly check-in, 30-day re-score, and 90-day re-score are what ensure the money actually stays moved.
Keep the local vault open in Obsidian for quick reference during execution.

Pause - I'll come back to this later

Keep the repo and local files intact. Snooze the cadence until the next strategic bet arrives.

Remove or snooze the weekly calendar check-in. Resume it when you pick up a new bet.
Keep the 30-day and 90-day re-scores booked for your current decision. Those still apply to the bet you just made.
Leave the GitHub repo and Linear project as-is. Both are on free tiers with no cost.
Leave the local set-the-bet/ folder in place. You will need it when the cadence fires.

Delete - I don't want to use this again

Remove all project resources. The completed brief and lessons learned are the artifacts worth saving separately if you want a record.

Delete the local vault folder. Open your terminal, navigate to the parent directory containing set-the-bet/, and run the command below.

macOS

rm -rf set-the-bet/

Windows

Remove-Item -Recurse -Force set-the-bet\

Delete the GitHub repo. In your repository on GitHub, go to Settings, scroll to the Danger Zone section at the bottom, and click Delete this repository. Confirm the deletion when prompted.
Delete the Linear project. In Linear, open the "Set the Bet" project, go to Settings, and select Delete project.
Remove all calendar entries. Delete the weekly check-in, the 30-day re-score, and the 90-day re-score events from your calendar app.

That's a Wrap!

Nice work. You just built a complete AI-facilitated strategic leadership drill, ran it on a real bet you own, and shipped the reusable system as infrastructure for how you make strategic calls from now on.

You've learned how to:

Build a complete AI-facilitated strategic leadership drill with four personas, eight timed phases, refusal rules, and a capstone scorecard that hard-fails any strategy without a real trade-off or a named human cost.
Produce a one-page decision brief on a real strategic choice you own, with all seven fields filled and scored at the capstone bar.
Ship the complete system as a tagged release in a private GitHub repo with Linear tracking, and book the 30 and 90-day cadence that checks whether the money actually moved and the team can still repeat why.
Secret Mission: ran the harder competing-bets variant, forced the kill decision between two defensible bets, and documented what it taught about trade-offs, trust, and the difference between leadership and management.

Ready to quiz yourself?