Ask HN: What's the next evolution of the GUI OS?

30 points by berkimbayev ↗ HN
Hi, in the past few years I was thinking about the evolution of OS based on GUI. Right now, we have all these windows and menus and lists. And you've probably noticed there is some uncomfortable limitation. For example, when you're working on some complex task, you can rarely feel the process flowing. You need to open different windows or different programs and go back and forth between them, spending time and mental resources. What can be a better solution to this?

What if we remove windows completely? Think about it - when you open a program, you don't actually need a separate window. What you need is functionality! So what if we remove windows and keep only the functionality. For example, you open a .doc file, but there is no separate window. You stay in the same "Desktop", but now you have an open file and functions to work with this file. And if you need some other functions, you open another program - and other functions appear on the desktop. So now you have two different programs working as one! And you don't need to arrange different programs and files on the screen. Because now you can have different programs and files working simultaneously as one.

I think that will be an incredible experience! And it will tremendously increase productivity and improve the workflow. If you're interested or have a comment, let's discuss below.

69 comments

[ 5.2 ms ] story [ 135 ms ] thread
I like this idea, but as a designer I foresee challenges in optimizing what is visible as you add new "functions" when needed. for example, editing a doc, then editing an image to add to the doc. Traditionally each function would live in an app with gui optimized to that function (Word vs Photoshop). Getting these functions to coexist when both are needed would get difficult to display, as image editing has a screen full of tools usually. Not saying it cant be done! I recall Microsoft had a similar idea with OLE (object linking and embedding), where the menus and tools would change depending on what the user clicks on within what is currently displayed. (See https://en.m.wikipedia.org/wiki/Object_Linking_and_Embedding )
Another example:

When I open two .doc files, I can only have two separate windows or tabs. But I don't need those windows or tabs. What I actually need is just two open files on the screen. And some Word functions. From UI perspective, it can be a "function bar" on top of those two files.

If I want to open an image file and Photoshop, currently I can only have a separate Photoshop window. What I actually need is just an open image file on the screen. And some Photoshop functions - a "function bar" on top of that file.

So the visual difference is that you now have 3 different open files and some "function bars" - all in one workspace. And you can arrange or scale your files in any way you want. And you can arrange "function bars" on the screen or hide them.

I don't understand what "files" means to you. If you have two multi-page documents, you have to somehow arrange them on the screen, and you need to show functions to edit/view/[...] either of them.

But, whatever you do, you need some kind of borders to understand where one document ends and the other begins, and you need some controls to change the view of each individual document (such as minimizing or maximizing it) and you need a button of some kind to close the document. Then, if we have two documents and floating bars that you can put wherever you want on the screen, it will be completely impossible to tell which bar controls which document (or displays the state of which document), so we should probably put the function bar somewhere in the document container. And, hey presto, we've reinvented the Window, because it turns out to be a really really useful and natural abstraction.

Okay, browsers, games, Excel, Calculator, and other type of programs obviously need some "window". But Calculator can be actualy made as a "function bar" that works with any other program.
One trend I'd like to see is more dense UIs. You can tell a lot of web and desktop GUIs are designed by developers with wide monitors because when you go to use those apps on a laptop with a small screen they are hard to use and you lose context from scrolling so much.
You are absolutely right. For example, I use laptop with Windows and 150% scaling. Otherwise the default scaling is 200% and everything looks so big.
I scale most websites to 150-200% because I like to fit the text I'm reading so that the ideal line length is approximately equivalent to my screen's width, and so that the things I need to click take up a significant portion of the screen. I'm a little short-sighted, but more significantly: I like to scope what I'm dealing with to an extreme degree.

Many websites break down when you scale them beyond. For example, news.ycombinator.com goes out of screen at >200% on my screen.

My terminal default font size is 20, but I often scale it above that when lines output is short enough for a period of time.

I love big output without fluff like task bars and clocks and other clutter.

What time is it? I don't know, I'll check my WRIST WATCH! :-)

A universal Undo API, with graphical indication of which actions can be added to the undo stack (e.g. file move), and which cannot (e.g. send email).
So you're saying you want to go back to MDI interfaces? I mean, we could, but it seems to have decisively lost the mindshare battle. I haven't seen any applications use it in ages.
I hope it's not transparent iPads like every sci-fi movie uses /s.
For the future, I'm anticipating vertical screens rather than horizontal ones. Laptops, for instance, you'll open up, and then rotate the screen so it's vertical. Big movie screens will also be vertical, requiring theaters to have much taller ceilings than they now have.

Why would you want this? Well, apparently, everyone these days likes watching movies and such on vertical screens, so that's the way the industry needs to move. I don't get it myself, but then again I'm not someone who watches TikTok videos.

I think you just need a bigger monitor. There are a lot of situations where multiple windows are the way to go. A designer needs to have the style guide PDF open while working with Photoshop, or an accountant needs to have its banking website on sight alongside the ERP. It might be possible for small tasks though.
Yes, there are situations where side-by-side comparison is invaluable.

Large laptop screens (>15") can be sufficient here, but two external monitors are priceless.

A headless OS came to mind immediately when I was reading your post. Let me expand: an OS which does provide a collection of components that can be used to build a UI but not the concept of windows and to make it connect to the hardware it provide the connecting functionality in a headless manner. App store transform into component store. Would be interesting to work on something like this.
Have you played with Plan9 the operating system and its windowing system?

or at least the Acme editor, which is also available for Mac and various Unix systems like Linux, FreeBSD etc.

https://github.com/9fans/plan9port

Plan9 (which I actually used for a few years) is caught in the anachronism trap of requiring a three-button mouse to effectively use rio.

It became unbearably obtuse to use on a laptop, and I didn’t like any of the workarounds, so I eventually upgraded my “terminal” to Linux and a custom LXDE setup.

Sounds quite similar to a terminal, no?
Insufferable answer: Emacs does this to some degree by overlaying "minor-modes" of function on top of one another, even when visiting the same file.

e.g. You have a TSV file which you might want to align visually and edit multiple fields simultaneously, so you enable `tsv-mode', and then you enable `multiple-cursors` and you have the desired editing functionality on that file.

But to answer your first question: I genuinely envision my ideal future desktop as being some smart-glasses kind of thing, where I can type by thinking. Right now, what my mind wants and what my hands want are often aligned, but my mind wanders. If I can type at the speed that I think, I have the impression that this will greatly increase my overall focus and mind-discipline.

(I can also envision employers forcing their workers to install intrusive mind-type monitoring apps, to ensure that they're actually working and not just drooling into space, but that's perhaps a separate discussion.)

I think, long term, apps will disappear and we will get APIs and algorithms that are controlled by an AI engine. Apps are data silos that prevent an AI from doing the things it could.

You should be able to tell the AI engine something like "as soon as the weather forecast predicts snow, order a snow blower from Amazon". The AI can only do this if it has access to weather data and an API from Amazon it can control to get you the snow blower. It should also be able to check your calendar, because you can't accept the delivery if you are on vacation.

Apps are silos that prevents the AI from doing these things. At best the AI could use the equivalent of screen scraping to use existing apps, but eventually all apps should be replaced by APIs that just extend the capabilities of the AI engine.

And what about GUIs? I think they will be pretty irrelevant. The AI engine should be able to render your data any way you want. Think Dall-E. There will certainly be designs to chose from, but you should be able to customize them any way you want ("remove that italic button, I never need it").

This is really wishful thinking.

>At best the AI could use the equivalent of screen scraping to use existing apps, but eventually all apps should be replaced by APIs that just extend the capabilities of the AI engine.

Why would companies want to expose APIs in this manner? Instead, they can force you to access them only through their app, and that way they can force you to sit through advertising, and otherwise control how you see the information and how you're allowed to interact with it. Open APIs cut them out of a lot of profits.

>You should be able to tell the AI engine something like "as soon as the weather forecast predicts snow, order a snow blower from Amazon".

This is crazy: this way, someone could just have their AI do these tasks in an automated way, without paying some company for the privilege.

Instead, what we need is an app that's the only way to access these APIs, and the app will give you ways to set up automation, but only in the ways the company wants you to, and poorly. Then they can charge you a monthly subscription fee for access to this automation app, and also embed annoying advertising in it.

> Why would companies want to expose APIs in this manner?

They don't. But why would record companies want to license their music to streaming services instead of selling CDs? If the AI is there, and it offers a superior experience to users, the companies that don't support it will disappear.

> this way, someone could just have their AI do these tasks in an automated way, without paying some company for the privilege.

You're assuming that APIs need to be free, but that doesn't need to be the case. Companies that provide data will need to charge for that, just like record companies charge Spotify when you stream music.

Companies that provide a service (Amazon, Uber etc) won't be happy that they lose marketing opportunities and brand recognition, but again, if the user experience is superior to the old app ways, those companies that don't adapt will die.

>If the AI is there, and it offers a superior experience to users, the companies that don't support it will disappear.

For that to happen, the AI needs to already be working. If the companies proactively prevent this from happening, then users won't have any basis for comparison. Closed APIs will prevent them from having the ability to program their own AI to do these things.

>You're assuming that APIs need to be free, but that doesn't need to be the case. Companies that provide data will need to charge for that, just like record companies charge Spotify when you stream music.

Sure, companies could require you to have a subscription to access the API, but they can make more money by controlling the experience instead of just giving you (paid) access to the API.

>Companies that provide a service (Amazon, Uber etc) won't be happy that they lose marketing opportunities and brand recognition,

Yes, this is exactly my point.

>but again, if the user experience is superior to the old app ways, those companies that don't adapt will die.

Users won't easily go backwards to proprietary access when they're used to (more) open access to something, but this isn't the case in this scenario: the users haven't had easy access to APIs before in a way that a regular person can just, for instance, give a voice command to their system saying "when the weather is like this, do these things with these various services", so they don't know what they're missing. The services already control the experience, and will want to keep it that way.

> For that to happen, the AI needs to already be working.

I don't think that the AI will be primarily developed to replace apps. I would rather expect Alexa or Google Assistant or Siri to start supporting arbitrary APIs.

> Closed APIs will prevent them from having the ability to program their own AI to do these things.

I would expect that a company like Google is able to find partners, but if not, there's always screen scraping.

> but they can make more money by controlling the experience instead of just giving you (paid) access to the API.

Not if people stop using the controlled experience. It just takes one competitor to offer the API.

>I would expect that a company like Google is able to find partners, but if not, there's always screen scraping.

They can shut that down with laws banning the practice. It might not stop someone working on a personal project, but it would prevent any significant company from offering any kind of paid product or service using screen-scraping. In fact, it might already be forbidden just by ToS.

Google or Apple could change their app store policies to either require allowing it or offer APIs.
> Apps are data silos that prevent an AI from doing the things it could.

If you're a company, that's not a bug, that's a feature. Data is your asset and releasing it for free is almost always a net loss.

It doesn't need to be free, but companies will need to adapt. I don't think there will be apps in 30 years.
Why would it suddenly be for free? One would probably pay for using data or API.

Actually one more way towards few companies controlling the internet. Because having an agreement with one supplier is more easy then contracts with many.

A voice interface makes much sense in a world were everybody wears augmented reality glasses. Manual input will be difficult and focussing on text or complicated graphics also.
I think it will be mixed, using both pointing device or touch screen, and voice. Some things are easier with a pointing device. For example, when browsing the web, I could rather touch or click on a link than saying it out loud. But other things are too complicated for a UI. Like saying "please use the photo I took of that dolphin last weekend as background" is much easier than finding the option to change the background, finding the dolphin photo and then selecting it.
> You should be able to tell the AI engine something like "as soon as the weather forecast predicts snow, order a snow blower from Amazon".

There is absolutely nothing in the above declaration that requires an AI.

I think the translation from human speech to something a computer can execute already qualifies as AI. But there are a lot of other things that need to be considered, if the AI aims to be the equivalent of a human assistant. It needs to determine how much snow requires a snow blower (a smart AI wouldn't buy for a couple of snow flakes), to guess what model of snow blower the user would want (which may involve checking the users' finances, brand preferences and it may ask how much space is available in the garage) and to determine what else needs to be checked, like vacations. It would also need to figure out the right time to order, because it should arrive before it is needed.
Those are all fair points - but if the AI "aims to be the equivalent of a human assistant", that requires interaction; a human assistant will ask question to understand the request if it is not clear. I think that the goal of having an AI "magically" figure these things out on its own is a fantasy.
Yes, it definitely needs interaction if the user did not define preferences in earlier conversations.
Back in the day, we used to have several monitors when developing web applications, mainly to mimic laptop use and desktop use. I thought that the dynamic resizing features in modern UI development would negate the need for these, but apparently not. Future of UI, or rather HCI, has to be voice or gesture. First company to make an intuitive interface and UX for the internet that everyone can use for their TV will be onto a winner.
Maybe. I suggest reading About Face : on Interaction Design if you’re interested in this topic OP.

The Affinity Suite has a nifty feature where, if you own the other packages, you seamlessly can edit a photo from Photo in Publisher for instance. Or a design in Designer from Photo etc. nothing loads, it just works.

I am writing half an OS, the front half. It displays a full GUI in the browser. The goal is native distribution over a network with full decentralization.

For example in Windows you can drag and drop file(s) between windows to move those file(s) between directories. In my application you can achieve all the same interaction which may move those file(s) to a different directory, different personal device, or different user. The experience is the same whether the application is running on Linux, Windows, or OSX.

In this application when you double click a file, in the browser to execute it, the OS default application associated with that file type opens just like in the native OS. If that file is on another personal device or another user it is copied to the local device for execution.

In order to minimize clutter I have a button to minimize all windows to a tray, but that is hardly original. Soon I will start working on a terminal emulator (also displaying in the browser) for remote terminal access to personal devices.

I know many people don’t like interfaces written in JavaScript but it is portable to things like Electron and Tauri. The test automation executes in less than 10 seconds on just the local device or just under 2 minutes for full peer to peer testing. Because this is all JavaScript there is a bunch of low level concerns I don’t have to reinvent, like accessibility, but the socket based decentralization does require new transmission handling and some original security work.

I need a separate window.

My WFH environment I use every day has three screens stitched together for 7680x1440. On the left third is my primary remote desktop app that connects me to work with a window for my old desktop PC. The middle third has 320pixels reserved on the right for context-sensitve notes, then splits the rest in half with the left half being my timesheeting/ticketing app and the right half being a decent plain text editor as a scratch pad / diary combo. On top of that I have a set of tabbed windows (using Groupy) from various apps like a file manager, browser, Slack, email, or command prompt. The right third has the left 320 pixels reserved for a music player (Winamp), VOIP softphone, and clipboard manager. The rest of that portion is another set of tabbed windows usually at least including a second remote desktop session to my VDI and another browser window. My browser and command prompt are tabbed. My file manager is dual-pane and tabbed with a preview pane.

What you're describing sounds like what was tried and failed with Windows 8.

It's possible what you describe could be achieved by AI compositing. I.e. The AI creates a central view of what it believes is the most relevant info for your current task. This doesn't require any changes to the programs themselves.

Longer term, I think voice + AI assistants can help with this. In programming speak, our UX will become more declarative and less imperative. Combined with AR/VR interfaces, this will see a big improvement in productivity.

> What if we remove windows completely? Think about it - when you open a program, you don't actually need a separate window. What you need is functionality! So what if we remove windows and keep only the functionality. For example, you open a .doc file, but there is no separate window. You stay in the same "Desktop", but now you have an open file and functions to work with this file. And if you need some other functions, you open another program - and other functions appear on the desktop. So now you have two different programs working as one! And you don't need to arrange different programs and files on the screen. Because now you can have different programs and files working simultaneously as one.

This idea of a "universal container" as strong similarities to Apple/IBMs "project pink" and OpenDoc [1] in the 90s that went on to become Taligent [2] that went on to become... not very much. There was also Microsoft's roughly concurrent efforts with OLE [3] and Windows NT Cairo - OLE was a stab at what might be described as componentising application functionality at the UI level, and Cairo had (iirc) the ambitious goal of replacing (augmenting?) the windows file system with a semantic data store with OLE/ActiveX as the UI layer.

OLE did kind of work, and actually made it into products like Word and Excel - though whether that was a good thing is debatable. The rest of it basically got canned, probably because it was too difficult to implement and/or sell. But I'm sure the influence lives on to some extent. I'm not aware of any serious attempts to go down this route again.

I still have some OpenDoc SDK beta CDs somewhere. It might be interesting to get them out and have a look.

Personally I think that "application" as a discrete concept provides a useful conceptual boundary [4] - although I will admit that this may just be that my brain is by now hard-wired for it.

[1] https://en.wikipedia.org/wiki/OpenDoc

[2] https://en.wikipedia.org/wiki/Taligent

[3] https://en.wikipedia.org/wiki/Object_Linking_and_Embedding

[4] Both for people using the application and for the grubby activity of selling it to them. Got to have a "thing" to sell.

This question smells like product research…but it’s intriguing so I’m bookmarking.

Fascinating responses thus far.

I think GUI has evolved too far, what I want is devolution, not back to what came before, but back to other basics.

We're in a terrible state, some odd and ugly place between native and online, where even native apps are picking the worst from online and online is denied access from the best of native.

I want an operating system GUI that dictates the available widgets to the program, and hands me control of their presentation, I don't want applications, online nor native to be able to dictate their own look, only for them to present their available functionality to the OS, and have the OS compose the screens for me, probably guided by my predefined preferences, sort of how HTML could have been if CSS hadn't ruined the day, sort of how native winforms programs used to be done.

A concrete example is things that can save a file, I don't want the program to dictate the gui for this, I want it to keep the OS updated "I have these documents open which have been changed since last save" or "There are these things which can be printed", and have the OS dictate GUI for activating that functionality.. It used to LOOK that way, for a lot of things, you'd go to the file menubar, and there you would open/save/print your file, this was easy to learn and very consistent..

Currently, the most consistent interface experience on my computer is bash in xterm, in elinks, websites look pretty consistent, if they're well designed.

But if I open firefox, chrome, gimp and kate, nothing looks or feels the same.. Then I open blender and a fifth kind of interface appears. No thanks.

So I sort of agree with you that it'd be nice if we did away with the window metaphor to some degree.. But I was also no fan of the way Mac had the "god bar" at top (which I believe Ubuntu did too) terrible idea..

For now, the least crappy desktop experience for me is i3, but it does not change anything fundamental, it's still program windows, free to do whatever they want.

But, for some things, the most futuristic version of the environment is the one that does not need a screen.. The tasks that are silently done for you without needing you to transfer any significant amount of information between you and the machine, or being able to do so on another level than a screen (brain machine interfaces and whatnot)

I want to generally echo this sentiment, for a slightly different reason.

I think the "dynamic-ism" of many interfaces has gone too far, therefore losing the predicability. Some interfaces try to guess what I am supposed to do next and it is correct maybe 60% of the cases.

But even when it is correct it actually slows me down because I am constantly second-guessing the interface and double-checking that what I am about to press is the one I actually wanted.

I wish I could rely much more on muscle memory and that things re-open where I left them. Like a window from a folder _always_ opening exactly where it was when I closed it previously. Or the homekit buttons on my home iOS pull-down screen not rearranging themselves randomly.

Edit: Another irritating example of this "trying to anticipate but keeps hampering" behaviour is the Outlook mobile app. When moving an email into a folder it will suggest three recently used folders and after a while re-arrange them after 3-4-5-something mails have been moved in a row.

I absolutely agree-

I want there to be places for things, a fairly rigid system so I can depend on things being in their place, not having to constantly "try to understand what's happening now"

Imagine if programs were more like plugins to the operating system, they'd register the stuff they could do, and the stuff they need to be done, and ultimately let the OS and user chose.. it could be that my text editor registers that it can edit text of ascii and utf-8 encoding.. it might register that it has plaintext files of that encoding available for printing, saving and viewing..

My browser can register that it can also display files of those encodings, and so I could effortlessly open a "viewer" in context of my text-editor, have it be my browser, and then when I edit the text-file, it is automatically shared (maybe even accessing same memory by pointer if I'm living dangerously) with the browser so I get live preview of my html.

The preview functionality in this case is not something that's explicitely supported by the text-editor, nor the browser.. It's just that since my OS already knows that the editor provides a source of text, and that the browser can view a source of text, then I can chose to open a browser perspective into the editor-source..

It might be that the operating system provides a "save file to disk" source, that I can then use to save the file from any program that provides a source of data.. It'd work the same way, instead of opening a browser as a perspective into my editor-text, I'd open a file destination into it, or I could open a LZ77 compression destination into it, which then opens a file destination, and my file is automatically compressed and saved, no program needs anything but tell the os what it can read and write..

To a certain extent, you're describing the KParts framework from KDE 3/4 times (I don't know if it still exists today). Many people remember Konqueror as a file manager and web browser, but it was actually a generic URL navigator that would load the right KParts component for whatever thing was being displayed. For an HTML file, the web browser component. For a directory, the file browser component. For a video, the media player component.

When I was developing Palapeli (a puzzle game that ships as part of KDE Games), I did a KParts version of it for fun in a few lines of code. That made it so that if you had a puzzle file stored in a folder, you could navigate to it in Konqueror and the game view would come up inside of Konqueror.

Back in the day, I dreamed of a fully document-oriented computing experience. Sadly the world has settled on the service-oriented experience of today, largely controlled by a small number of gigantic fiefdoms.

Oh, I do remember that Konqueror had this really cool way with URLs, but I didn't realize the depth of it at the time!
When it comes to programs that work with files (such as Word or Adobe Illustrator), the window is just a border around the file. It doesn't have any other purpose. I think of it as barriers to the workflow. I have to arrange windows on the screen. What I actually want is to arrange files. And have functionality from different programs - all on the same screen (stacked like those function bars in Word or Adobe Illustrator).
And you will clearly see the difference between "function bars" of different programs. For example, Microsoft and Adobe have different UI language. You will have different files open on the screen, different function bars around, and no window borders. No barriers to work.
I like the Office app where you have Word, Excel, and PowerPoint in one program. But still you need to open files in different windows/tabs. Is it possible to make files open in the same window/tab? So that you can work on two different files without any borders. This will be very close to real-world experience where you don't have any windows/borders to work on two different papers. What if we make a file itself as a window?
You need the right OS underneath. Ideally something based on persistent objects like OS/400 and with the networking aspect of Plan9 so that local and remote resources can be treated the same.

The GUI would need a similarly smart approach and given an OO persistance layer I'd say that applications would just be objects that composed other objects - the document object would be your wordprocessor and it would be good at arranging and presenting other types of objects into 2d representations at A4 size.

The web browser would presumably just be a superset of the document object or might not even be needed at all as a web page is just another kind of document.

You'd still need a "fresh sheet of paper" concept but also a history so you could go back and search for what you were looking at/working on previously. Also different contexts - like all the stuff you need for one bit of work and another context for some other kind or play/work. I don't think windows are the worst way of doing this and I'm certain that you'll end up wanting to look at different things side by side or whatever and that will become windows by another name but there might be subtle improvements.

An object oriented desktop instead of an app oriented desktop sounds like a good idea to me. Show me things and groups of things instead of apps. A person, a message, a video.
Microsoft did this in the early 90s . It was called OLE (Object Linking and Embedding). It was a nightmare of different UIs for different purposes all fighting over the part they edited