One of the most rampant things I've noticed with AI trying to design product UX is that is LOVES progressive disclosure. So much you don't see in the UI at first, hidden behind certain clicks or logic, and that only show themselves the more you click around and mess with it.
2
70
I was using @Replit for something at work today and noticed an interesting design decision. Each type of agent action has its own icon, and the icon appears every time the agent performs that action, even if that means repeating the same icon over and over within a single response. I’ve attached a few examples from my session.
The first time I saw all the duplicates side by side, I thought it was glitching. I’m not sure how much value all the repeats add. They could probably get away with showing just one icon per action type in each response, or even none at all. Just the logo animation they have plus "Thinking..." could be enough without adding the current clutter or confusion, plus all the aforementioned icons are shown in the expandable action history anyway.
49
I generally agree with this sentiment. Reminds me of how it's critical to avoid asking leading questions while doing user research so that you don't get a tainted answer. Once an idea is planted, it becomes so much harder to think totally outside of that box.
However, going to LLMs for things like copy and mockups definitely has its time and place. Sometimes if I'm completely stumped on a design problem or just not happy with where my own ideas have gotten me, going to Claude Design or Figma Make and asking for 5-10 unique approaches to something can be just the spark I need to get the creative juices flowing. Sometimes I end up using something similar to a concept the AI provided, but sometimes I riff on it as a starting point and end up with something unique and far better than where I'd been.
In short, it's important not to jump straight to asking AI to do/solve everything for you, but it's also definitely not something you should shun and shut out completely.
i'm finding it increasingly important to not ask LLM-s for copy / visual mocks because the basic patterns and cliches will poison my brain and make it harder to design / write beyond the average
1
123
I was a little iffy at first on how the Claude Code desktop app’s sidebar peeks in on hover, disappears when you move away, and stays pinned once you actually click it. The more I’ve used it, though, the more I’ve come to really like it as an interaction pattern. Claude made fairly quick work of building something similar for a project I’m working on at OpenComp. I still need to get the tooltip delays just right so they don't clash with the sidebar's movement, but I like it so far.
1
151
Love this but have never understood the reasoning behind the alignments pointed to here. The indentation of these elements in particular should align with the left side of the "Product" section title. In the current state it feels like the Issues and Projects icons are floating in an odd in-between.
Replying to @karrisaarinen @thekitze
there should be vertical/horizontal rhythm. Either be left-right positioning, or centering.
So first top bar should be aligned or vertically centered with the traffic lights
the left side panel should have padding and either left icons aligned or centered
then generally you want either round or rectangle buttons or fields. you might mix if there is some special case
you can also send apple HIG to your agent to try to follow the rules developer.apple.com/design/h…
I was ripping on a project with Claude Code but then started craving a coffee. Decided to try Remote Control on my CC session (I know, I’m late to this feature) while I made a coffee run and omg it’s amazing 🤩🔥
1
103
Something I've been experimenting with on a recent project: Using stripped-down, verb-based action phrases as what's visible by default, then progressive disclosure to reveal the actual controls once intent has been shown by toggling on a respective switch.
Daniel Solberg retweeted
it's so crazy how for years we as an industry went on about empathy maps, persona designs, user journeys, wireframes, only for everyone to become obsessed with the idea of one-shotting interfaces with artificial intelligence.
Even as Claude Code has become absolutely integral to how I build, brainstorming and iteration in Figma has remained a critical part of my workflow (shoutout to the infinite canvas ✊😤).
For this sidebar project, my Figma artboard was messy and spanned dozens of different ideas. I explored tiny, minor changes and bigger "crazy" ideas before finally settling on a direction that I then wanted to put into code.
Once that version was built out in code, it became a process of refinement to get it just right or to address anything I might have overlooked on the static Figma canvas (movement, etc).
I definitely don't think Figma is dead. In fact it's actually quite fun to bounce between Figma and Claude Code as needed, and it's a new type of flow state that I've been really enjoying.
Yesterday we launched an overhauled navigation UI at OpenComp, designed entirely by me, and also built entirely by me using Claude Code.
Our previous nav bar was a dark marine color (a dark shade of one of our brand colors) that felt heavy, and utilized horizontal nav tabs that felt dated. Having the old nav along the top also took up a lot of valuable vertical real estate, especially for being a SaaS app where maximizing screen real estate is critical. It was long past time for us to move into the future and adopt a collapsible sidebar instead. Now, with the new sidebar implemented, both vertical AND horizontal space can be maximized.
I intentionally went with muted colors for the sidebar so that colors throughout the app's main pages can draw users' attention the most. I wanted the navigation to fall back and be there when it's needed rather than competing to be the star of the show.
Hierarchy was interesting in the sidebar because the user can access different, what we call "scenarios" (think different copies of the same general environment) in the app, each of which has the same set of nav tabs accessible within them. So each scenario 'encompasses' the navigation, but users aren't changing scenarios very often, so the scenario switcher doesn't need to command a lot of attention. I solved for this by making the scenario switcher sit above the nav tabs and be slightly taller than the nav tabs themselves. These two design decisions give it some dominance in the hierarchy through placement and size while keeping those cues subtle. There's also intentionally no BG color on the selector until it's hovered so it doesn't draw attention or act as a constant visual anchor.
Some other things I really like about the new sidebar:
• The OC logo and the icons stay in their exact positions between both the expanded and collapsed states. This makes closing and opening the sidebar feel nice in the sense that those elements aren't shifting around while the movement is happening, even while other elements appear and disappear.
• Not pictured here, but an 'upgrade' card sits nicely in the lower left corner above the Settings and Need Help buttons for users who are on free accounts or free trials.
• With the transitions between the open and closed states, it was my first real foray into transition timing. The scenario selector, Settings, and Need Help buttons all appear smoothly as the sidebar expands (rather than instantly), which was a cool detail to focus in on. I still need a lot of practice with transitions and movement, but I'm glad to have dipped my toes in a little bit.
I actually worked on this project 4-5 months ago, but it wasn't released until now for other reasons. It was a huge confidence boost as my first "real project" using Claude Code, and one that has now ended up in our customers' hands. My creativity and productivity have really taken off ever since, and I'm now deep in some even more ambitious OpenComp projects with Claude that I'll share more about soon.
To be able to hammer out this new sidebar with no dev support needed outside of the code reviews was so exciting and such a cool feeling.
I can't wait to keep building! 🚀
3
4
214
Yesterday we launched an overhauled navigation UI at OpenComp, designed entirely by me, and also built entirely by me using Claude Code.
Our previous nav bar was a dark marine color (a dark shade of one of our brand colors) that felt heavy, and utilized horizontal nav tabs that felt dated. Having the old nav along the top also took up a lot of valuable vertical real estate, especially for being a SaaS app where maximizing screen real estate is critical. It was long past time for us to move into the future and adopt a collapsible sidebar instead. Now, with the new sidebar implemented, both vertical AND horizontal space can be maximized.
I intentionally went with muted colors for the sidebar so that colors throughout the app's main pages can draw users' attention the most. I wanted the navigation to fall back and be there when it's needed rather than competing to be the star of the show.
Hierarchy was interesting in the sidebar because the user can access different, what we call "scenarios" (think different copies of the same general environment) in the app, each of which has the same set of nav tabs accessible within them. So each scenario 'encompasses' the navigation, but users aren't changing scenarios very often, so the scenario switcher doesn't need to command a lot of attention. I solved for this by making the scenario switcher sit above the nav tabs and be slightly taller than the nav tabs themselves. These two design decisions give it some dominance in the hierarchy through placement and size while keeping those cues subtle. There's also intentionally no BG color on the selector until it's hovered so it doesn't draw attention or act as a constant visual anchor.
Some other things I really like about the new sidebar:
• The OC logo and the icons stay in their exact positions between both the expanded and collapsed states. This makes closing and opening the sidebar feel nice in the sense that those elements aren't shifting around while the movement is happening, even while other elements appear and disappear.
• Not pictured here, but an 'upgrade' card sits nicely in the lower left corner above the Settings and Need Help buttons for users who are on free accounts or free trials.
• With the transitions between the open and closed states, it was my first real foray into transition timing. The scenario selector, Settings, and Need Help buttons all appear smoothly as the sidebar expands (rather than instantly), which was a cool detail to focus in on. I still need a lot of practice with transitions and movement, but I'm glad to have dipped my toes in a little bit.
I actually worked on this project 4-5 months ago, but it wasn't released until now for other reasons. It was a huge confidence boost as my first "real project" using Claude Code, and one that has now ended up in our customers' hands. My creativity and productivity have really taken off ever since, and I'm now deep in some even more ambitious OpenComp projects with Claude that I'll share more about soon.
To be able to hammer out this new sidebar with no dev support needed outside of the code reviews was so exciting and such a cool feeling.
I can't wait to keep building! 🚀
How is that in 2026 CSS still has no support for controlling dash length or gap when border-style: dashed?? This is wild to me with how powerful CSS is in so many other ways.
1
45
Daniel Solberg retweeted
Dear designer, don’t panic
Use a barbell strategy for designing with AI
Zoom way out. Work hard by hand to understand and architect systems / stories / the big idea
Zoom way in. Pick moments to obsess unreasonably and invent clever twists in corners of the work where the human touch is irreplaceable: microinteractions / prose / animation / visuals / brand
In the expansive middle, welcome the company of your obsequious robot underlings, and orchestrate them broadly with your exquisite taste and considerable suspicion
I’m basically building all frontend for work projects myself now with Claude. It’s been an interesting dance, though, to figure out when engineering needs to get involved to lay the proper foundations for me to layer on top of (and whether that needs to happen first or can happen in tandem along the way). How are other designers handling this process?
1
43
Daniel Solberg retweeted
As a 20+ year design engineer, i can say emphatically that AI enables designers to build great PROTOTYPES or small apps.
When @dhh talks about hand-written code being dead...I believe him. I believe this is mourning engineering as it was. But I don't think hand-written code === engineering. I think it was just how we did it.
An engineer thinks in systems, and ways that designers don't often think. I don't think coding is dead. I think it's changed. I value so much my engineers and I witness their talent on the daily, PAIRED with AI as well.
Long live engineering.
One of the coolest parts of Claude Code is having your localhost window up while it's working, and seeing the page or feature you're working on literally transform and come to life in real time as it cooks 🤩
1
61















