Liquid (Gl)ass and the future of technical aesthetics

https://www.macrumors.com/guide/ios-26-beta-3-liquid-glass-changes/
(God forgive me for posting an OP link to macrumors)

For those living under a rock, Apple recently released a new design system they call liquid glass. It is present in the iOS 26 beta, which I've been using as an exercise. The beta v1 had unadulterated liquid glass, but they have since been walking it back and in beta v3 many of the occurrences have been nerfed to something more like a naive opacity. My highly controversial opinion after continued usage: backing off of the original vision is a big mistake, because full-send liquid glass and the general paradigm it represents is the future. And boy is the future gonna be good.

As a matter of historical analogue, Apple introduced the aqua interface in 2000, which had a very similar visual identity to liquid glass. Since the very early days of computing we have known that blur is computationally expensive, so various hacks were culturally embedded like the aforementioned naive opacity and/or clever use of pngs that look like artifacts you would see in a bonafide object that was made of glass or tinted acrylic. In video games we have lighting approximations like Phong for highlights, etc.

Jonathan Blow observed that although hardware has been improving at an exponential rate, GUI software looks/works much like it did in the 90s (i.e. bad/poorly). We are in a pretty deep potential well of HTML, CSS, and JS and have been for decades. To first approximation, video games have been the only alternative domain to the web for technical aesthetic expression, and their absolute spiritual supremacy is undeniable. They have the mandate of heaven.

Over the past few years we've started to see some interesting things happen with the web, like WASM and WebGL. A few companies have taken advantage of these, the most notable is probably Figma (IPOing soon). Their stuff all runs in the browser and yet feels like a video game because it is written like a video game and takes advantage of hardware like a video game. In general the trend is that despite web devs being mostly incompetent idiots, there is a clade of priests advancing the technical aesthetics of video game development into the web, and they are being rewarded handsomely by Gnon for their superior interpretation of His will.

We now return to liquid glass, aka the Linux ricer's wet dream. The shaders Apple is using, assuredly implemented in their proprietary Metal Shading Language (MSL), are glorious. Dynamic chromatic aberration, specularity, physically accurate refraction, I mean it's a proper scientific testament to Beauty. In beta v1 it was on nearly every control surface of the entire operating system, and it didn't even nuke battery life all that much. Remarkable, thank you very much TSMC. Vertex and fragment shaders decide every pixel of every frame buffer in video games, and they are deciding an increasing number of pixels in every frame buffer of all modern computational interfaces.

We must thank God for that. I have written enough CSS and HTML to satisfy one hundred lives. Please, no more. The graphics frontier is reopening after a long hiatus, and once again Aryan Intelligence (AI) may dominate Computer—this time around with Rust and GLSL. I humbly conclude this screed with a few glorious references to shader-as-art:

https://x.com/XorDev/status/1943750402387648778
https://x.com/XorDev/status/1937221326327963657
https://x.com/zachlieberman/status/1933969965788393885/photo/1
https://x.com/zachlieberman/status/1929626460853358711/photo/1
https://x.com/YoheiNishitsuji/status/1931100986220945776
https://x.com/YoheiNishitsuji/status/1889106921330016668

>>3629
Liquid glass is definitely cool. Too bad they weren't able to make it more usable. You're right that video-game style UI rendering with custom geometry, shaders etc is the future. I have only done a comparatively small amount of CSS and html and it already makes me want to kill myself. We're stuck in an insane and ugly equilibrium and it does feel like the way out is custom high-compute graphical simulation techniques as in videogames.

Sofiechan has a unique aesthetic but it's firmly in the previous paradigm, or even the one before that. I wonder what something like it would look like projected into the shader future. It would be an incredibly fun project at the least. So let's see:

We'll start with the abstraction of 2.5d or 3d space as a lens into high-dimensional cyberspace. Great aesthetic innovation always starts as skeumorphism to connect to existing cognitive structure and then abstracts and purifies into something new (see greek columns evolving from the form of trees). Text has thoroughly undergone that process, but computer UI is continually open to innovation as graphics techniques advance. There was a moment in around y2k where videogame UIs almost took us down this path, but videogames were still low prestige, it was too expensive, too difficult, and the previous era of apple aesthetic was still pulling us into aestheticized flat blank rectangular purity. By 2015 with windows 10 and material design that reached its peak IMO. The last 10 years are a blur and it feels like we've gone nowhere. But now gpu based graphics are way more powerful and accessible and the aesthetic possibilities of the blank rectangle are aesthetically exhausted so maybe its time to go back to the custom spatial skeumorphism track again?

>>3631
So let's imagine future sofiechan in 3d cyberspace. I imagine it as a city one can explore. Tags are neighborhoods. Threads or posts are buildings. Position is calculated by linguistic topical proximity. This could be styled with infinite character like some pastel anime seaside ancient greek polis of philosophers. Imagine statues of athena and perseus looking over us.

Or maybe its a more abstract datascape in high dimensions that is merely projected onto 3d, with maybe a dominant up-down dimension representing the activity-deadness priority we currently communicate with front page presence and bump order. Maybe you can rotate the whole thing through other dimensions to pull out different aspects as you surf through cyberspace. This is only worth it if those different dimensions are really intuitive. Otherwise, 2 principle components of proximity is plenty. Or maybe it rotates to flatten onto the *local* principle components as you zoom in to a particular area or tag.

This invites us to think about "background" and the nature of the space itself. It's got a color gradient skybox at least. The eclipse logo is the sun. Maybe there is some kind of landscape or horizon in the distance. The central mesa or "land"mass holds the city itself. What is the nature of that? I don't know. Maybe it flies like Columbia from bioshock. This is all in need of more exploration.

Very little of this kind of thing has been seriously explored. It's like we knew it was supposed to go this way (read some pre-y2k scifi!) but in practice it was too kitsch and awkward and we ended up preferring this spiral down into the flat minimal rectangular mind prison. If that's where we need to be, give me a text layout engine and let's get rid of all the extraneous ornament. This is the way books went, and sofiechan so far has done a good job of playing to this abstract utilitarian minimalism of the written word. But it feels like a betrayal of the possibilities of the medium, and flat text sucks for navigation of what is in fact space. If it's in fact space, let's start embellishing and mapping it as space.

>>3633
The right abstraction for text itself, as we have figured out over many hundreds of years, is glyphs on a page. The page is a thin, flat featureless surface on which text and images are laid out. It ideally has minimal physicality of its own except maybe the ability to be lit, or some of its own light, and the slight texture or color or its material. Via images, the page can contain or be a portal into spaces, or can contain other pages. Magazines and infographic posters take this to the limit of richness. Pages are embedded in a larger higher-dimensional space. As a special case, the page may also be a control panel with touchlike interaction. This is more or less where we are with the web etc. The attack of the GPUs comes from two ends: what space is the page embedded in? And what material is the page made of that in can exist in that space and we can interact with it?

Apple's answer to the latter is simple: drops of liquid glass. Liquid glass gives everything a uniform intuitive and beautiful physicality to finally free the digital page from being conceptually a rectangular array of pixels. I like the idea, but liquid glass itself is questionable for a variety of reasons. It still seems optimized for novelty and impressiveness rather than intuitive utility. It would be very luxurious if it worked, but it doesn't quite work. I think they went too hard on the skeumorphic fidelity instead of imagining the material they really needed, which at the very least would have been translucent rather than transparent. Skeumorphic fidelity was itself a substitute for artistic direction. "make it behave like liquid glass" is an easy thing to communicate, with very few degrees of freedom for mis-creativity. Control of degrees of freedom has always been the problem in UI design. "Make it behave like this imaginary material that is actually the right thing for an interactive page medium for all these specific cases" has too many degrees of freedom. But it's close, and they've sent the strong signal that it's time to explore. Refinement, abstraction, and perfection of the idea will come with time.

If I were a competitor and wanted to one-up them or play the new paradigm in my own direction, I would go with "frozen cloud": panels of translucent ice that grow and recede as needed to fit the desired rough shape, but have an optical texture somewhere between the transparency of ice and the opacity of paper, like a panel of heavily frosted glass. It allows some light to pass through but is readably opaque. Crucially, it has extreme potential for high fidelity simulation of naturalistic ornament in the dendritic crystal growth around the edges. Finally something better than the corner radius! My quick idea for the material metaphor for images and text is they are sortof painted on with "magic smoke" that evaporates in and out of existence as needed. Dark mode and light mode can be naturally functions of how the ice gets dark with ambient lighting and the magic smoke lettering turns on to glow in the dark. Interactions with the magic smoke can send light through the ice panel for the tactile magic animation feel of the whole thing.

Place the frozen cloud ice crystal pages in a cloud-scape and you've got a harmonious paradigm for both the surrounding space and material details. This is fun to think about and I bet it could be done well if one had a budget for a team of genius shader artists. It will be interesting to see how apple and others actually handle it and where the paradigm goes.