A friend started her business as a sophrologist a year ago. A former journalist, she knows how to write, and she brings back superb photos from her travels in a wheelchair. She wants a site to bring her offer together properly: Instagram has its limits, and prospects regularly ask her for her website.
I told her, a little curtly, to start on her own. Instead of hours of discussion, three Word files, three zips and three emails, a rough prototype built with Claude: good old pompous slop, and 70% of the work. She did it. One HTML file of 405 KB, 7 pages, 3 languages, 20 photos.
Slop, made to order
I wrote in July that the best brief is a draft. This one has the form I have been removing from my generator since June, slop, and it bothers no one: it goes in the bin, her words and her photos stay. The prototype shows me what she wants, in her own words, across 7 pages.
The remaining 30%
They are mine: the art direction, the site map, the UI, the UX, the overall experience. On 26 October, Claude transcribed the 7 pages into the CMS I am building for the web, social media and print, same sections, same order. Then I sculpted. The logo was 40 pixels in a corner; it now opens the home page full screen. The press and the testimonials were tucked away in two menu pages: they move up to the home page, as entries that can also feed a post or a carousel. Email and WhatsApp, drawn like the social networks, become two big buttons, so that people write to her before they follow her.
Her prototype could have gone online as it was: an HTML file takes a minute to host. A useful site costs next to nothing in 2026. A site finished at 70% can cost a lot in lost business.
On the 27th, a V2 fits in 4 pages instead of 7, and a button switches from one to the other. Once a version is chosen, the other will disappear, and the button with it.



