Design the page in a thread before you build it

Argue the layout out in conversation, section by section, and leave with a mockup and a spec your coding tool can follow.

The expensive version of this is building the page, looking at it, and rebuilding it. Doing the argument in a thread costs a few generations and the thing you walk out with is a spec rather than a commit you have to unpick.

Why this will not look generic: The page follows an argument you made out loud, section by section, rather than the model's idea of what a landing page contains, which is the thing every generated page defaults to.

What you walk away with

• A section-by-section layout you actually reasoned through
• A full-page mockup generated from it
• A written spec a coding tool can implement against
• The prompt set, so the next page in the same system matches

This one happens in one conversation

You bring: What the page is for and who lands on it. No design yet, that is the point.

Open with: I am building a landing page for X aimed at Y. Before any images, argue with me about what sections it needs and in what order, then we will generate it.

What the thread holds on to:

• The argument you had about section order, so the mockup follows a decision rather than a template
• Every variant it already generated, so 'more like the second hero' resolves to a real thing
• Your saved style, if you have one, so the page is in your system from the first draft

Then you can just ask:

• Drop the testimonial band and show me what the page looks like without it
• What is this page missing that a sceptical buyer would want?
• Write the spec as Tailwind classes and a section outline
• Now generate the social card for it in the same style

How it works

1. Argue it out, then generate (In Claude) - The brief above is the step. Resist generating until the section order is settled.
2. Build the full page here - The thread settles the structure; this renders the whole page rather than a hero crop. Full-resolution output lands in your library.

Why do this in a thread rather than a prompt box?

Because the valuable part is the argument, and a prompt box has nowhere to keep one. You get a page on the first try and no way to say why it is wrong.

• The reasoning survives the regeneration. When you drop a section and regenerate, the thread still knows why the other sections are in that order. A fresh prompt does not, so every iteration quietly relitigates decisions you already made.
• You leave with a spec, not just a picture. The output that matters to a developer is the section outline and the class names, not the image. Ask for both at the end and the thread writes them from the decisions rather than describing the picture back to you.
• The next page matches this one. Save the prompt set and the second page starts from the system rather than from scratch, which is the difference between a site and a pile of pages.

Best before you have written any markup, which is the window where changing your mind is still free.

Cost and time

This workflow runs at from 10 credits + about 1 minute. Your result carries automatically from each step into the next, so there is no re-uploading between tools. Reverses are billed from your reverse meter and generation steps from your AI Credits, both bundled in every pay-as-you-go credit pack.

Browse all guided workflows or all individual tools. Pay as you go, no subscription, AI Credits never expire.