It's Monday. Here's what changed in AI this week, and what it means for your design work:

This week was all about Limora's onboarding. The launch brought in a lot of traffic, but traffic isn't the goal - conversion is. So I went back through the whole flow to find where people were dropping off.

First, I emailed everyone who signed up but didn't pick a plan. Limora doesn't have a free tier on purpose. It's a premium tool with premium generations, and a free plan just fills it with people who were never going to take it seriously.

Then I fixed the biggest friction point. Instead of asking you to upload your brand kit by hand, we built custom scripts that pull the brand kit straight from any website. Way less work, especially on mobile, where digging up a logo file is a pain. We also changed what you generate before the paywall: instead of a mobile logo variation, you now get a custom hero image and hero section based on your project reference. Two days in, conversion is already up.

On the personal side, I took a one-day trip to Barcelona. Loved it. I'll be back this summer for a week or two to see what it's like to actually work there - the vibe in summer is hard to beat.

This is the first AI model built for people who judge their work with their eyes.

Anthropic pulled Fable 5 offline for six days after a government-forced shutdown. As of June 18, it's back. The model launched on June 9 and did something no Anthropic model had done before: it took the vision lead from OpenAI. That sounds like a benchmark footnote. For designers, it's the whole game.

Most AI models read your design. Fable 5 looks at it. You can hand it a screenshot of an interface and it'll tell you what's off - spacing, hierarchy, contrast, balance - then rebuild it in code that actually matches. It has taste, not just templates.

For two years, AI has been great at generating and terrible at judging. You'd get ten layouts and still have to be the one with the eye. A model that can genuinely see design changes that. Screenshot-to-UI stops being a party trick and becomes a real workflow: screenshot a site you like, hand it over, get a clean build back that holds up. The designers who win with this won't be the ones generating the most - they'll be the ones who know what good looks like and can now direct a model that does too.

One catch: the restored version comes with new limits (nationality-based access, tighter safety filters) and falls back to Opus 4.8 more often than before. Worth knowing if it suddenly feels less sharp.

Next time you're stuck on a layout, screenshot your draft and a reference you admire, drop both into Claude running Fable 5, and ask it to spot the gaps and rebuild yours to match. Start there.

You design a screen, then spend the next hour rebuilding it as a prototype someone can actually click. Magic Patterns skips that step.

It's an AI prototyping tool that turns a prompt into working UI - and not generic UI. Import your design system and it generates screens that match your existing styling, so prototypes look like your product, not a template. Designers and engineers can edit the same file together, and you can pull from a catalog of over a million community designs.

This is how you stop getting AI slop and start getting work that looks actually designed.

This turns Claude into a full UX strategist instead of just a code generator.

Ask it to design a flow and it handles error states, touch targets, and WCAG contrast on its own.

Built by Anthropic, this lets you describe a poster, graphic, or piece of artwork and have Claude design it for you.

It exports straight to a production-ready PNG, no Figma or Canva needed.

Claude scaffolds a complete editor that exports App Store and Google Play screenshots at every required size.

Device mockups, captions, and light/dark variants are all handled for you.

💌 Know a designer who should be using AI smarter?
Forward them this email. Or just send them to logiaweb.net to join.

See you next Monday,

Say hi on X, Instagram or YouTube