Unicorn Club note: no newsletter for the next couple of weeks.
I am taking time to mess with the format and move it onto the site. I do not want it to live only as a weekly email.
Same idea: better product work, sharper judgement, more useful examples. Just changing shape.
For AI UI work, the boring bit might be the part teams need most.
Pick one repeated thing, like a card or empty state, and write down the real component, the states it needs, what the tool can infer, and the fake versions that would pass a quick glance.
One AI design-tool test I'd use now:
Does this screen use the real component, states and rules, or has it made a convincing impression of them?
Looks close is useful, but if the tool can't show its receipt I'd still treat it as a sketch.
The Design System Guide looked at 158 public systems.
89% had code examples, but only 37% had usage guidance, 21% covered accessibility, and 13% included content guidance.
If AI tools are reading the docs now, that missing context becomes product behaviour.
Ugly AI UI is easy enough to reject.
The annoying one is the screen that feels familiar enough that you stop checking where it came from.
That is where I want a receipt: real component, real states, and the bits the tool guessed.
The old design-system drift was someone making a one-off button.
The new version is an AI tool making twenty screens that feel close enough to pass a quick review.
That's the bit this week's newsletter is about.
158 public design systems reviewed:
89% had code examples
37% had usage guidance
21% covered accessibility
13% included content guidance
That was already awkward for humans. It gets worse when an AI tool is trying to read the system.
Tomorrow's newsletter is about the bit of AI design work that worries me more than the obviously broken stuff.
The screen that looks like it used your design system, but can't prove it.
The new Unicorn Club format got its first proper write-up today, and it’s all positive.
Big relief honestly. I was confident the change made sense, but it’s still nice when someone else sees what you were aiming for.
Tomorrow's newsletter is the first proper go at the new format.
One product/interface thing I've been digging into, what shaped the thinking, and the bit I think is worth taking into your own work.
AI can remove a lot of boring work, which is good, obviously.
But if it also removes the part of the flow that showed the mistake or gave someone a final check, I'm not sure that's actually simpler.
That's the thinking behind tomorrow's newsletter.
The prompt box isn't really the bit I care about.
The interesting bit is what the product shows after AI has guessed something, changed something, skipped something, or filled something in for you.
That's what tomorrow's newsletter is about.