For Jeremy Adas · April 2026

From idea to published.
Your Claude playbook.

A four-stage path with fifteen copy-paste prompts. Built for a builder who hasn't called himself one yet — on a Chromebook, with no jargon, in a single afternoon.

4 Stages
15 Prompts
One Chromebook
Zero Excuses

You're already building things on Perplexity from your phone.
That's not "not technical." That's a builder who hasn't named himself one yet.

So this isn't a Claude tutorial. It's a system. The same four stages every real maker walks through — idea → PRD → iterate → published. And fifteen prompts that move you from one stage to the next without thinking about what to type.

The path

Four stages. In order. No skipping.

95% of people jump from "I have an idea" straight to "let me start building." That's why most projects die in a tab. The PRD stage in the middle is the unlock.

01

Idea

What if

02

PRD

What is exactly being built

03

Iterate

Make it actually work

04

Published

Ship it

Your stack

Each tool in its lane.

You don't have to pick Claude OR Perplexity. They do different jobs. Use both. Use Copilot at work, within Fidelity's policy. Stop scattering.

Primary · Personal

Claude

Thinking, writing, building. Idea → PRD → Iterate → Publish. This whole playbook lives here. Web-based — runs perfectly on your Chromebook.

Research · Personal

Perplexity

Keep using it. "What's the current state of X?" Computer-use agent for go-do-this-on-a-website tasks. Already strong for you.

Work-Only

Microsoft Copilot

Inside Fidelity policy. Don't mix work data into your personal Claude account. Different account, different rules — keep them separated.

Casual

GPT

Search replacement, fun stuff. Light queries. Nothing wrong with it; just doesn't earn the primary slot.

The stages

What each stage actually is.

Three things every stage has: the goal, the time it takes, and the trap that kills 95% of projects. Know the trap; avoid the trap.

Stage 01 · Idea

Take fuzzy → make it sayable.

~30 minutes · Prompts 1, 2, 3

Get the idea out of your head and into one sentence a stranger could understand. Let Claude poke holes. Cut to the smallest useful version.

The trap Building something nobody wants. If you can't say in one sentence who it's for and what problem it solves, you're not done with Stage 1.

Stage 02 · PRD

Write the contract with yourself.

~45 minutes · Prompts 4, 5, 6

One page. Who it's for. What it does. What "done" looks like. What you're explicitly NOT building. Especially the "not building" part — that's where projects die.

The trap Skipping this stage because it feels like paperwork. The PRD is the difference between a project that ships and a permanent open browser tab.

Stage 03 · Iterate

Build it. Look at it. Fix one thing. Repeat.

~3–5 hours across multiple sessions · Prompts 7–11

Get v0 in front of your eyes. Diagnose before you patch. Fix one thing at a time. Polish only after it functions. Stop when the PRD says you're done.

The trap "Almost done" syndrome. You'll be tempted to keep adding features. Don't. Ship the PRD. Add features in v2.

Stage 04 · Published

Get it out of Claude. Tell one human.

~1 hour · Prompts 12, 13, 14, 15

Simplest hosting that works. A README so future-you remembers. Tell at least one person it exists. A 5-minute retro to make the next project faster.

The trap Treating "Published" as optional. If it's not in the world, it doesn't exist. Even a saved Claude Project counts — but you have to tell someone.

The 15 prompts

Copy. Paste. Move forward.

Every prompt has a clear job. Use them in order the first time through. By project three, you'll know which to skip and which to remix.

01
The Idea Dump
Externalize the fuzzy idea. End with a one-sentence project statement.
Idea
02
The Hole-Poker
Pressure-test the idea. Three reasons it might not work + the one biggest risk.
Idea
03
The Smallest Useful Version
Cut to the bone. One core feature. What's left out for now.
Idea
04
The PRD Writer
One-page Product Requirements Doc using a fixed template. The contract.
PRD
05
The Out-of-Scope Sharpener
Add five more "we are NOT building this in v1" items. Aggressively.
PRD
06
The Success Criteria Test
Could you demonstrate each criterion with a screenshot? If not, rewrite.
PRD
07
The First Build
Hand over the PRD. Get v0. Simplest tech that meets the requirements.
Iterate
08
The "What's Wrong" Pass
Diagnose before patching. Hypothesis + two tests + smallest possible fix.
Iterate
09
The Targeted Fix
One specific change. No refactoring. No "while we're at it." Surgical.
Iterate
10
The Polish Pass
Once it works, make it look good. Typography, spacing, color, mobile.
Iterate
11
The Honest Review
Compare against the PRD. Ready to ship, or one more iteration?
Iterate
12
The Hosting Plan
Three hosting options ranked from simplest. Chromebook-friendly. Free or cheap.
Publish
13
The README
One-page documentation future-you will thank present-you for writing.
Publish
14
The Share Post
Three drafts: a text to a friend, a LinkedIn post, a curious-friend note. Send one.
Publish
15
The Retro
5-minute retrospective. Save lessons for the next project's system prompt.
Publish

Things to leave at the door

A few myths that have been slowing you down.

These are real beliefs you walked in with. Each one is wrong in a way that's costing you projects. Worth knowing before we start.

You think
Reality
"Claude needs a powerful local computer."
Claude.ai is web-based. Your Chromebook is fine. Identical experience to Perplexity.
"Claude is more technical than Perplexity."
The chat interface is the same shape. Claude is just better at thinking, writing, and building.
"I'm not technical."
You build apps on your phone with computer-use agents. That's more technical than 95% of office workers.
"I have to pick Claude OR Perplexity."
Different lanes. Different jobs. Use both. Smart users carry a small toolkit, not one hammer.

Pick a first project. Run Prompt 1. Go.

The first project is the slowest. The third feels automatic. By the fifth, you won't need this guide — you'll have built your own version, tuned to how you think.

Email Reu — let's go