WHAT THIS PHASE PRODUCES
You leave with
- A market context note: who exists, what they’re doing, what nobody’s doing yet
- A P0 scope you can actually validate this loop
- A visual direction tight enough to brief a prototype
THE MENTAL MODEL
How to think about this phase
Plan is where most founders over-spend. They mistake the phase for “write the PRD”, fan out into a six-week research project, and arrive at Test with a roadmap they’re too invested in to throw away.
The point of Plan is not comprehensiveness; it is sufficiency. You need enough market context to know which trapdoors you’re walking past, and enough scope discipline to leave with a P0 you can build and test inside one loop.
If you find yourself adding features to the P0 because they’re “easy”, stop. Easy is a vibe-coding trap. The P0 must isolate one variable: the single thing your hypothesis is actually about. Anything else is noise that will make the test result un-attributable.
Visual direction belongs here too, not as polished design but as the shared mental image that lets you brief a prototype without re-explaining the bet five times. A wireframe sketch and three reference URLs beat a Figma file you’ll throw away.
PRACTICE PROMPTS
Paste these into plain Claude or ChatGPT
Adaptive market scan
Run a focused market scan for this hypothesis: "<paste yours>" Return: - Five existing tools or workflows the named customer uses today, with the specific job each one currently does for them. - For each, name the exact friction my hypothesis claims to remove, and the evidence (quotes, reviews, public posts) that the friction is real. - One unsolved adjacent job that nobody has shipped a clean answer for. Be specific. "Automation" is not an answer. Do not give me a feature comparison table. I need to understand the workarounds, not the products.
Cut to a P0 you can validate this loop
Help me scope a P0 for this hypothesis: "<paste yours>" The P0 must: - Isolate the single variable my hypothesis claims will move - Be buildable in 5–10 hours of focused work using AI tools - Be testable in 8–12 customer-discovery sessions inside two weeks Walk me through three candidate P0s, then argue for the one that strips the most without losing falsifiability. Tell me what to defer and why deferring it doesn’t weaken the test.
Visual direction in one pass
Describe a visual direction for the P0 we just scoped. Treat this as a brief, not a design. I want: - Three reference products or screens it should feel adjacent to (and one it should NOT feel like) - The single most important on-screen artifact (the one element a user will read first) - The shortest possible label set the prototype needs Do not propose colour, type, or a component library. Stop short of design choices the prototype builder should own.
Paste a prompt into Claude or ChatGPT, then replace the bracketed placeholders with your own work. Free, no signup required.
FREE AGENTS
Run a free Plan agent
adaptive market scan
Market Gap Finder
Describe your idea in one to three sentences. Returns a copy-paste-ready research prompt for Google NotebookLM that maps industry factors, surfaces where competitors over-invest or under-deliver, and identifies the least-contested space to test in.
define the customer
ICP Strategist
Walk from a broad demographic to a specific, named customer segment. Returns a profile covering what they're trying to accomplish, the workaround they're stuck with today, and what would make them switch. Ready to guide your first customer interviews.
working backwards
Future Press Release
Writes the Working Backwards PR FAQ from your hypothesis: launch-ready press release, sharp customer and problem framing, internal and external FAQs. Forces strategy decisions before any code.
scope the P0
PRD Generator
Paste your target customer profile, user journeys, or feature scope. Returns a structured product spec formatted for direct use with AI coding tools. Defines exactly what to build.
COMMON TRAPS
What goes wrong in this phase
- Building a feature comparison table. Founders always do this. It produces nothing actionable.
- Scoping a P0 that tests three things at once. Whatever the result, you won’t know which variable moved it.
- Treating Plan as the place to lock the roadmap. The roadmap belongs after a loop closes, not before.
- Rationalising scope creep with “while we’re in there”. Every extra surface adds a confound.
WHAT GOOD LOOKS LIKE
A worked example
SETUP
From the Idea example: the “first-draft PRD” bet for Series-B PMs.
OUTPUT
After Plan: “P0 = a single web form where a PM pastes a problem statement and receives, within 60 seconds, a structured outline plus the specific clarifying questions still missing from their statement. No saving, no auth, no team features. We test the trade: would they actually paste their next real problem in.”