The 3-prompt method | How AI-led builders write to Claude (and never settle)
How AI-led builders write to Claude (and never settle for the first output) 🪺
TL;DR: Vibe coders write 1 prompt and ship what comes back. AI-led builders write 3. The first prompt is almost never the right one. The third one almost always is.
Here’s the rhythm I use for every build — landing pages, apps, branding, anything you’d otherwise hire someone for.
⌁
Prompt 1 — Vague
Write what you’d say to a friend. One sentence. No structure.
Example: “Build me a landing page for a candle business.”
What you get: A generic page. Could be selling anything. Boilerplate copy. → Not shippable.
Why it matters: See what the AI assumes when you give it nothing. The vague prompt is diagnostic — it shows you exactly what context is missing.
⌁
Prompt 2 — Directed
Add the context the first prompt missed. Audience, goal, vibe, exclusions.
Example: “Build a landing page for Ember & Oak, a small-batch soy candle brand for people who want a calmer evening. Audience: 30–50, design-conscious. The page should feel quiet, not loud. One CTA: join the waitlist.”
What you get: A real-feeling page. Audience-aware tone. → Almost shippable.
Why it matters: This closes 80% of the gap between what you want and what the AI guesses. Most people stop here. AI-led builders write one more.
⌁
Prompt 3 — Specific
Leave nothing to chance. Brand colors, typography, layout, hero copy, hard nos.
Example: “Same brief. Colors: cream + deep teal + warm wood. Serif headlines, sans body. Hero: one candle photo, full-bleed, copy: ‘A candle for the quiet hours.’ Do not use stock-photo people. Do not say ‘transform’ or ‘luxury.’”
What you get: The page you actually meant. → Shippable.
Why it matters: This is the prompt you actually build from. The specificity is the work.
⌁
The lesson:
The work isn’t in the building. It’s in the prompting. AI didn’t change that — it just made it more obvious.
Each prompt teaches you what’s missing from the next one. That’s the whole method.
⌁
Want the copy-paste PDF? Comment “PROMPT” and I’ll DM you the full 4-page guide — the method, the candle example walked all the way through, and a template you can use on any build.
Follow @unexpectednest 🪺 — Sundays = prompts. Wednesdays = builds.
#indiehacker #aitools #buildwithai #aiforfounders #promptengineering
In my experience working with AI tools like Claude, I’ve found the 3-prompt method to be a game-changer for getting the exact output you want. Often, I see users make the mistake of giving just one vague prompt and expecting perfect results. This usually leads to generic or uninspired content that can’t be shipped or used directly. The power of this method lies in its iterative nature. Starting with a simple, conversational prompt helps you observe how the AI interprets minimal instructions. It’s like taking the AI’s temperature to understand its baseline assumptions. From there, you add important context about the audience, tone, and purpose, which closes most of the gap between AI output and your vision. But the real magic happens with the third prompt — the one where every detail is spelled out clearly, including brand colors, voice preferences, and what to avoid. This level of specificity transforms AI-generated content from generic to tailored and professional. I’ve applied this approach across multiple projects—landing pages, app designs, marketing copy—and the improvement is always tangible. For example, when creating a landing page for a handmade candle brand, my initial vague prompt gave me something bland and uninspired. Adding the audience and mood details brought the text closer to what customers would resonate with, while the final detailed prompt nailed the brand’s vibe and visuals perfectly. This method pushes you to think more critically about your AI instructions and teaches you what you might be unintentionally leaving out. It also aligns well with the best practices of prompt engineering, which emphasizes clarity, context, and constraints to help AI perform optimally. In conclusion, don’t settle for the first AI output. Embrace the rhythm of prompt iteration—vague, directed, and specific—to craft compelling, on-brand, and shippable content every time you build with Claude or similar AI tools. This approach not only enhances the quality of AI results but also trains you to communicate more effectively with machine intelligence.
Prompt