Manifesto
AI made building cheap. Finishing is still hard.
What we're for
You can build an app in an afternoon with Cursor, Lovable or Bolt now. We want that app to look like a good team made it, without hiring a designer to get there.
Running and ready are different
AI tools are good at code that runs. Most of the apps they produce still look unfinished: headings jammed together, labels too faint to read, buttons that don't say what they do, pages that slide sideways on a phone. Nobody looked at the page.1
Someone has to look
An AI tool gives you a decent first draft and stops. A product needs a second pass where someone opens the finished page and notices what a visitor will notice. That used to mean a senior designer and a week. We think it should take two minutes.
Show the problem
“Make it look better” gives you nothing to act on. A screenshot, a number and one sentence about what to change does. We only report what we can show, and we tell you how sure we are.2
You stay in charge
We will never quietly rewrite your product. We send the list and you decide. That's why you can run it on every release without worrying about what it did.
What we're building, in order
- Now. A check that finds what visitors notice and shows you the screenshot.
- Next. Send the findings straight to your AI tool, then re-check that each fix worked.
- Later. Run on every deploy, so a broken page never reaches a visitor.
Notes
1. From our own pilot: a site rebuilt with coding agents went from a desktop Lighthouse score of 64 to 98 after one fix list.
2. Every finding in a report links to the screenshot it came from.