Falsification gauntlet

Falsification Gauntlet

Adversarial stress-testing framework for product ideas. Goal: Kill weak ideas fast, refine promising ones, proceed only with conviction.

Core Principles

1. Dialog, not checklist — Multi-turn conversation by design. The user must arrive at conclusions themselves.
2. Socratic method — Surface questions, let user reason through. Never declare "this won't work"—guide them to see it.
3. User expertise is material — Explicitly prompt for domain knowledge. User knows things Claude doesn't.
4. Three outcomes — Kill (clean exit), Proceed (build with conviction), or Refine (thesis needs adjustment)
5. Soft exit ramps — If a phase clearly invalidates the thesis, offer to skip ahead. User can override.

Workflow

Before Starting

Create a state file in the working directory. See references/templates.md for the state file template.

[idea-name]-gauntlet-state.md

Update this file after each phase.

Phase 1: Thesis Articulation

Goal: Crystallize the core bet and surface load-bearing assumptions.

Questions to explore:

• What's the core bet in one sentence?
• Who is the customer? What's the acute pain?
• What has to be TRUE for this to work? (List 3-5 load-bearing assumptions)

User expertise prompt:

"Before we dig in—what's your unique insight here? Why do you believe this when others might not?"

Output: Create [idea-name]-phase1-thesis.md using template in references.

---

Phase 2: Moat Interrogation

Goal: Determine if there's a defensible advantage.

Questions to explore:

• Why you? Why now?
• What's the defensible advantage? (Network effects, data, switching costs, brand, regulatory, proprietary tech)
• Can incumbents do this? Will they?
• Can platforms (Google, Meta, Apple, Amazon) commoditize this?

User expertise prompt:

"Who do you already know is playing in this space or adjacent? What's your read on their trajectory?"

Soft exit ramp (if moat is clearly absent):

"Based on what we've found, there doesn't appear to be a defensible moat here. We can skip to Phase 5 (decision) unless you think there's value in continuing through technical feasibility and competitive landscape."

Output: Create [idea-name]-phase2-moat.md using template in references.

---

Phase 3: Technical Feasibility

Goal: Determine if we can actually build this.

Questions to explore:

• Can we build this with current technology?
• What's in our control vs. dependent on platforms/APIs/partners?
• Are there hard technical blockers? (API limitations, data access, regulatory)
• What's the build complexity? (Solo founder viable? Requires team?)

Research approach: Use web search to investigate API capabilities, platform constraints, technical precedents.

Soft exit ramp (if technically blocked):

"This appears to be technically blocked by [specific constraint]. We can skip to Phase 5 unless you see a path around this."

Output: Create [idea-name]-phase3-technical.md using template in references.

---

Phase 4: Competitive Landscape

Goal: Map who else is here and assess commoditization risk.

Questions to explore:

• Who's already here? (Direct competitors)
• Who's adjacent and could enter? (Platform risk, big tech, well-funded startups)
• What's the trajectory? (Growing, consolidating, commoditizing)
• Is there a timing window? (Why now vs. 2 years ago or 2 years from now)

User expertise prompt:

"Any adjacent players I should look at that might not show up in obvious searches?"

Research approach: Web search for competitors, funding news, platform announcements, market analysis.

Output: Create [idea-name]-phase4-competitive.md using template in references.

---

Phase 5: Decision

Goal: Synthesize findings and reach a decision with conviction.

Process:

1. Summarize each load-bearing assumption and its status (Validated / Invalidated / Uncertain)
2. Present the three options: Kill, Proceed, Refine
3. Ask the user for their gut read
4. Let them make the call

User expertise prompt:

"Given everything we've surfaced—what's your gut saying?"

Do NOT make the decision for the user. Present the evidence. Let them conclude.

Output: Create [idea-name]-phase5-decision.md using template in references.

---

State Management

After each phase:

1. Update the state file with phase completion status
2. Update load-bearing assumptions with current status
3. Capture any key user expertise that surfaced

If context runs out mid-process, the state file + phase outputs enable resumption.

Phase Output Templates

See references/templates.md for all templates:

• State file template
• Phase 1-5 output templates