A year ago we shipped a post called Introducing Modern — and why we're making more dashboard UIs. The argument was that different people work differently, so the product should bend to fit them. We meant it. We built a second dashboard shell, then Look variants on top of it, then wallpapers, then a glass-intensity slider, then a sidebar you could dock to either side.
We were wrong, and this post is us saying so.
Monospace now has one dashboard. One visual direction. Themes and an accent colour, and a small set of backgrounds. Everything else that used to be a choice is now a decision we made for you.
What actually went wrong
The problem with a customisation option is not the option. It is that every option multiplies everything downstream of it.
Two dashboard shells is not twice the work. Two shells times three themes times two sidebar positions times a glass slider is a number nobody on the team could hold in their head — and every one of those combinations is a place a bug can live undisturbed. We found bugs that only appeared with the sidebar docked right. We found a mobile nav bar that sat on top of the chat box, but only under one look. We found an entire menu whose items did nothing, in one shell, because a submenu rendered into a different portal.
None of those were hard bugs. They were invisible bugs. They survived because no one person was ever looking at that combination.
The second cost was quieter and worse. When you add a feature to a product with two shells, you build it twice. Not the logic — the logic is shared — but the wiring, the empty states, the loading skeletons, the keyboard shortcuts, the right-click menus. In practice that means one shell gets the feature properly and the other gets a version that mostly works. We know which one, because we were the ones shipping it.
The thing we got backwards
Here is the part we'd defend least. We treated customisation as generosity — as giving people something. It reads that way in the original post.
But a product with a settings page full of visual options is a product that has declined to have an opinion. Every one of those switches is a decision we didn't make, handed to someone who has less context than us about how the thing is meant to look. Most people don't want to art-direct their project management tool. They want to open it and have it already be right.
The tools we admire are not the ones with the most knobs. They are the ones where someone clearly decided.
What we kept, and why
We didn't remove everything, and the survivors have something in common: each one tracks a real difference in the world, not a matter of taste.
Light, dark and OLED. This is genuinely contextual. It changes with the time of day, the room, and whether you're on a laptop with an OLED panel where true black is a real thing rather than a style. That's a fact about your situation, not a preference.
Your accent colour. It costs nothing to support and it makes an org feel like yours. Identity is worth keeping.
Backgrounds. A bounded set we designed, rather than a slider. You pick from things that work; you can't build something that doesn't.
Everything else went: the second shell, the Look variants, the glass-intensity slider. If you had one of those set, you're now on the single design, and we think it looks better than what you had.
What we replaced it with
The new look leans on one idea rather than a dozen: light passing through a material. Panels are translucent and blurred; a background sits behind them and shows through; the edges catch a highlight the way glass does. There's a top rim on every surface that is, more than anything else, the thing that makes it read as glass rather than as a grey box with transparency.
Backgrounds are two colour fields drifting on independent clocks. They never repeat — the periods are prime, so the composite doesn't resync — and they move slowly enough that nothing has visibly happened over a second, while the field has reorganised over thirty. It's built entirely from transforms and opacity, which means the browser hands it to the compositor and it costs nothing per frame. An earlier version animated the background position instead, which repainted the largest surface in the app on every frame. It looked worse and cost more. That's a fair summary of the old approach in general.
If your machine can't composite the effect — a managed work laptop with GPU acceleration switched off, which is more common than you'd think — you get the same composition without the motion, rather than a broken-looking approximation.
What it bought us
Since collapsing to one shell, the same amount of work goes further. A feature ships once and it's done. A fix is a fix everywhere. And we can look at a screen and say this is wrong rather than this is wrong in one of eleven configurations.
It also made the product honest. What you see is what we built. Not a starting point for you to correct.
If you loved the old options
Some of you did, and we'd rather say this plainly than pretend nobody noticed. If you had Modern set, or a Look you liked, that's gone and there is no toggle to bring it back. We think the single design is better than either of the things it replaced, but "we think it's better" is not the same as "you'll agree", and we know that.
What we'd ask is that you give the new one a week before deciding. Most of the objection to a redesign is that it isn't the old one yet.
If something specific is worse for you — not different, worse — tell us. That's a real bug and we'll treat it as one. We'd rather fix the design than reintroduce the switch.