Build for one person, one tension, one dream, one first step.
Before you write a page, build an agent, or ship a flow — name the human on the other side, the friction they arrived with, the world they wish were true, and the single positive action that moves them one step closer. Make that action easier than any other.
Intended action
Before writing, name the idea to instill and the first action it must cause.
Consequence
The page stops decorating the idea and starts moving someone into a better loop.
Most agents and pages fail not because the output was wrong, but because the builder never named who it was for or what first step would feel obviously right.
Profile · Tension · Dream · First Action · Cascade
The cascade lens inside stage 4 of The Loop.
Answer four questions out loud — then write.
These four answers ARE the brief. Output naming, consequence chains, kill signals — all downstream of getting these right.
Who is the reader, at what moment?
A specific person walking in with a specific worry — never 'users,' never 'people who might.' Role + arrival moment is the whole of the answer.
What friction did they arrive carrying?
The fear, the unmet desire, the moment something stopped working. If you cannot name what hurts when they open the page, you cannot design the relief.
What does their perfect world look like?
The state they would describe if asked 'if this just worked, what would be true?' The dream is the emotional engine — it makes the outcome matter and the first action attractive.
What small action moves them toward it?
The single positive next step — verb-led, low-cost, hard to do wrong. Name the wrong action they are most likely to take instead, then design the slope so the right one rolls downhill.
The cascade — 1° direct → 5° structural — earns its place AFTER these four. It stress-tests that the first action actually leads somewhere. Pick a profile below to see the discipline applied.
Three builders, three first steps.
Pick one. Read the four-question discipline applied end-to-end. The cascade unfolds inline as evidence the first action actually leads somewhere. The wrong action sits below, collapsed — available, not competing.
Owner / Principal
Wants the leverage of agents without losing the line of sight that protected the business when it was just them.
A business that runs while I sleep, with receipts I trust enough to stop watching every step.
Authorise the first bounded delegation
Reads agent.json. Sets a spend threshold. Enables exactly one declared capability. Watches the receipt land.
The owner moves from guarded curiosity to controlled trust because delegation is bounded and receipted — not surrendered.
First receipt arrives within the hour. Spend stayed inside the threshold. The next decision is bigger than the last.
Agent executes the declared task. Receipt emitted to IntentTrace.
Owner reviews the receipt. Trusts the trail. Raises the threshold.
More delegation. Attention shifts from doing to reviewing receipts.
Team learns: agents can be trusted because receipts exist, not because humans watch.
Organisation operates at agent scale. Human effort concentrates on strategy, not execution.
Dream first. Reality backwards.
The Better Practice package starts with the room people want to believe in, then works backwards into the artifacts that make investment, joining, or hosting rational. Pick an arrival to see the discipline applied.
Investor / First Believer
Wants the dream without underwriting a fantasy. Has funded vibes before and learned the hard way.
Capital that buys evidence — demand, repeatability, facilitation, pricing, and stop rules — before any scale call gets made.
Fund a 90-day reality Loop
Approves a small staged ask for a room, three practice protocols, session receipts, and Monday review.
Capital buys evidence: demand, repeatability, facilitation, pricing, and stop rules before scale.
30 / 60 / 90 day receipts you can read in one sitting. A named kill switch if the room doesn't pull.
The team runs the 30-day dogfood and emits receipts for every session.
Return, reflect, refer, and repeat signals show whether the room has pull.
The first external cohort tests whether the protocol works beyond the founder.
Belief shifts from founder charisma to a receipted practice loop.
The next investment decision is based on evidence, not vibe.
Investor / First Believer
Outcome signal
30 / 60 / 90 day receipts show return, reflection, referral, repeatability, and a paid ask tested.
Kill signal
No paid offer, no returning cohort, or no facilitator repeat test by day 90.
Operator / Facilitator
Outcome signal
A joiner can run the first three sessions from the protocol pack with Monday review evidence.
Kill signal
The work still requires founder improvisation after three attempts.
Partner Host / Community Buyer
Outcome signal
The host sees repeat attendance, useful participant notes, and a clear next cohort decision.
Kill signal
Warm event feedback without return behaviour or a buyer-owned next action.
Run the exercise. Then write.
The four questions feed the consequence chain — which is the Direct Action Contract made operational. Walk the steps in order; do not skip ahead.
Name the profile. Role + the moment they arrive + the tension they brought. A specific person at a specific moment — never 'people who might.'
Name the dream. What would they describe if asked 'if this just worked, what would be true?' Loud and specific. This is the emotional engine.
Name the first positive action they should take, and the reward they get for taking it. Verb-led. Low-cost. Hard to do wrong.
Name the wrong action they are most likely to take instead. Both shape the slope — design the right one to roll downhill.
Trace the five-order consequence chain for the right action. If the cascade doesn't compound into something structural, the first action wasn't the right one — go back to question 4.
Close with a CopyablePrompt that asks the reader to do question 4 themselves. The prompt is the first action — not a summary.
Put this to work
Answer the four questions before you write a word.
For the page designer / builderCopy this prompt. Paste into Claude, ChatGPT, or any AI assistant. The page context is already loaded — send it and get analysis tailored to your role.
I am designing a page for a business analysis, agent surface, journey, or venture. Before I write any content, I want to answer the four questions that determine whether the page causes the right action — or just looks busy. QUESTION 1 — Profile. Who is the reader, and at what specific moment do they arrive at this page? Name role + arrival moment in one line. Never "users," never "people who might." A specific person walking in with a specific worry. QUESTION 2 — Tension. What friction did they arrive carrying? What is hurting, missing, or worrying them at the moment they opened this page? One sentence. QUESTION 3 — Dream. If everything just worked, what would they describe as true? Write the line they would say out loud. This is the emotional engine of the page — make it the loudest line on the card. QUESTION 4 — First action + reward. What single positive action moves them one step toward the dream? Verb-led, low-cost, hard to do wrong. What reward do they get for taking it? And what is the most likely WRONG action they will take instead if the page fails? Then — only then — trace the 5-order consequence chain for the first action: 1° direct — what immediately results 2° downstream — what that first result causes 3° system — what that downstream effect triggers in the broader system 4° cultural — how repeated instances change beliefs and norms 5° structural — what the system looks like after this action compounds over time If the cascade doesn't compound into something structural, the first action wasn't the right one — go back to question 4 and pick a smaller, sharper move. Finally, write the single sentence the CopyablePrompt at the close of MY page must ask the reader to do. Not a summary. The first action itself — handed to them in a form their own assistant can run.