Here it is: https://adamsilver.io/blog/native-html-components-dont-guarantee-good-ux/
14.03.2026 18:01
👍 0
🔁 0
💬 0
📌 0
Here it is: https://adamsilver.io/blog/native-html-components-dont-guarantee-good-ux/
Native doesn't mean good.
Select, date inputs, file upload. They work. They're often terrible. The native element is a starting point.
Slowly coming around on Liquid Glass more than I expected to.
Accessibility questions are still real and there are quirks that need sorting, but overall I'm kind of vibing with it 🤷
What's everyone else thinking?
https://cutlefish.substack.com/p/tbm-409-minimally-viable-consistency
Ideas that spread without being mandated. That's the consistency worth having.
The rest is overhead.
Read it: https://bradfrost.com/blog/post/real-time-ui/
AI generating to a design system spec, not from scratch.
That's the framing that changes it from interesting demo to actual tool.
Here it is: https://web.dev/blog/baseline-navigation-api
Navigation API hit Baseline. Took a while.
I've been wiring up click listeners and patching history for years. Nice to have this native.
Here it is: https://piccalil.li/blog/an-in-depth-guide-to-customising-lists-with-css/
::marker is one of those things I assumed I understood until I tried to actually style it.
Three Stack Overflow tabs and an MDN page later.
Popover for overlays. Dialog for true modals.
I've defaulted to dialog for everything and then fought the accessibility model the whole way. Pick the right one before you build.
Automated scanners catch roughly 30% of accessibility issues.
The rest shows up when you keyboard through it. Or actually use a screen reader.
We keep telling people to check AI output. But we haven't written down what good output actually looks like.
So the check is "does this seem fine?" And it seems fine until it doesn't.
loading="lazy" for video and audio is in Chrome Canary.
Every media-heavy page I've built needed an intersection observer setup to fake this. Sometimes you just want the platform to do the thing.
Sometimes the paste bubble on iOS takes so long that I'll tap the field again to retry.
It shows up right as I do that. My tap makes it go away.
Rinse and repeat…
You're not slow. You keep rebuilding from scratch.
New tool every few months. Never quite long enough for anything to stick.
Compounding needs time. The problem was never the tool.
This week's Unicorn Club is about a CSS property that makes clip-path hacks mostly unnecessary.
Also: when design system contribution models break down, and why you should run postmortems after your successes as much as your failures.
Out Wednesday.
Thanks to our sponsor BugHerd
Pixelstorm uses BugHerd to speed up website feedback and launch with confidence.
No login for clients, point-click-comment, screenshots + technical details captured, tasks created for the team.
Read more go.unicornclub.dev/bugherd
This week's newsletter is out.
Design system enforcement, MCP tools for Figma, and tracking where research recommendations actually go.
unicornclub.dev/newsletters/...
Watched someone do 200+ variables in about 2 seconds with an MCP tool and i'm not sure what i've been doing with my time.
The time isn't even the interesting part. It's that you stop thinking about the mechanics and stay on the actual design decisions.
newsletter.baselinedesign.com/the-mcp-tool...
The planning cycle is where it breaks. "Attach to a high-priority initiative" is the right ask for a feature team.
For a design system team it just means macro thinking gets squeezed out by whoever needed something specific last sprint.
The pattern I keep seeing is that the AI output becomes the brief.
So you end up designing around a generated wireframe rather than the actual problem. The decision got made, just not by anyone who owns the outcome...
This week's newsletter is about the gap between what a design system says should happen and what actually ships
Someone builds a modal with no close button, QA passes, a user finds it. the system had the rule. nobody was enforcing it.
Out wednesday. unicornclub.dev
Thanks to our Sponsor - BugHerd
Website feedback with less back-and-forth. Clients point, click, comment.
BugHerd captures screenshots + technical details and turns it into tasks.
Free trial go.unicornclub.dev/bugherd
This week’s newsletter is about the risky moment when something looks done but won’t survive real users.
Three moves:
- Make the first minute feel safe, not clever
- Decide what’s disposable, then design for it
- Slow the noise, not the shipping, with release notes
unicornclub.dev/newsletters/...
Most AI UX mistakes are tone.
If the it feels like a colleague, users expect colleague-level judgement. They will overtrust it, then stop when it fails.
Design it like a tool:
- fewer pleasantries
- clearer boundaries
- explicit review and rollback
- ownership that’s visible
You win some, you lose some 😉
Status looks green until release week.
Replace your next update with a 5-line reality scan:
Shipped / Learned / Risk watch / Indicator / Decision ask (yes/no).
Makes risk routine and decisions explicit.
Use for biweekly updates and release notes.
More templates in the newsletter: unicornclub.dev