I don't understand how these type of projects are still tried and get any traction... anyone who has tried them will 100% know it won't go beyond a happy path demo. If they want to seriously use/publish the app beyond playing around, it'll require weeks of iteration via AI, which will cost you an arm and a leg in tokens.
Founder of Raycast here so obviously biased but you’ll be surprised. You get a working app one-shotted pretty much all the time. Sure if it is something more complex you might need a few more prompts. Just to give you some examples on what we’ve seen:
- Our support team runs on Glaze apps to review Raycast extensions. It connects to GitHub, checks out code locally, gets realtime updates and so on.
- The sound agency build a functioning synthesiser for the launch video. It works even with MIDI.
- We’re about to cancel a team-wide subscription and replace it with a Glaze app.
Not everything is possible yet and sure more complex things need more prompts but you’ll be surprised what Glaze is capable of already. It’s day one…
This sounds promising. If I may take advantage of you being here, what language does it write in? Does it build genuine native apps (Cocoa, WinAPI or WinUI, etc) or Electron?
The FAQ was light on technical details. But I am someone keen to read all the technical details :)
I haven't used v0 or replit before, I have the same feelings as you. But I've been thinking about building macOS apps for my personal use for a long time now. Also I'm a long time Raycast user. I have a bias here, so I've joined the waitlist, I can't be sure until I try, right?
My first thought was, "So, Replit and ilk?", seems they expected that comparison:
> How is Glaze different from Lovable, Replit, or v0?
> Those tools build for the browser. Glaze builds for your desktop. That means your apps can access your file system, your camera, keyboard shortcuts, menu bar integration, and background processes. Things a web app can’t do. It’s a different category entirely.
Pretty sure modern web apps can do all of those (sans menu bar). (If anything they do background processes better since you can send a very long task off to a server and shut off your computer, come back later and pick up where you left off.)
Also, as others mentioned, this just seems like Claude Code with extra steps, unless they managed to nail some sort of design standard enforcement they feel is better than what most people can get out of it.
The quick publishing is kind of nice, but it immediately made me think it would be more interesting to have a way to quickly remix other people's creations, similar to the Figma Community tab: you can take someone else's work, break it apart to see how it works, then tweak it how you want it.
I took a few shots at building desktop apps with Tauri, Wails and Electron using Claude Code, and the results were not very good at all. In fact, they were by far the worst results I've gotten with the tool. I can easily clone one of my boilerplate repos in Rails, or Django and prompt away, and the results are consistently good, as in, functional MVP in a few hours. This was never the case for the desktop tools I mentioned.
This looks like a highly specialized tool for desktop that actually works. I watch the demo and I am assuming the apps are actually made with some kind of technology a la Tauri, or Electron, thus making the apps cross-platform.
I don't think we are anywhere near a tool like this for native, but that's a lost battle anyway.
I’ve had a totally different experience. I’ve coded 3 different Tauri apps and 1 Wails app with Claude Code and it was some of the easiest work I’ve done with AI assisted coding. That said, the local features that Rust is handling in the Tauri app is not anything heavy, just moving files around, some regex matching, and some SQLite stuff. All of the headache I had in these apps was the React frontends and Node issues. The Rust features all worked pretty much first try every time.
> I don't think we are anywhere near a tool like this for native, but that's a lost battle anyway.
I hope it's not a lost battle, tbh. I was hoping with AI & Vibe Coding we'd see sort of a resurgence of native first desktop apps, but so far it's just all been a continuation of the web app & web tech hegemony.
Maybe not for Windows as their native GUI story is a lost cause now, but for sure macOS and I had hopes of it leading to a renaissance of desktop linux apps in GTK instead of electron, but that (the Linux) community seems to be hostile to any AI generated code at all for now.
I have been seeing more and more native desktop apps in the past few months (octarine for instance), but most of them would've honestly been better off as web-apps, or at least a polished electron app.
> seems to be hostile to any AI generated code at all for now.
Because the majority of vibe-coded apps are low effort.
Well, to be fair, I do have an experience working on a Windows Forms app from scratch. App connects to a very specific scanner via customs drivers and makes use of a remote API for data tasks. The app works, it's stable, but I'm not going to lie, AI assisted coding for this particular stack does require a very large amount of nurturing, it is just not the same experience you get with web apps. Nevertheless, it did it.
Makes sense. There's plenty of freely available code and data online for using web tech. Any number of free online bootcamps spawned in the mid 2010s are full of "Become a React developer in 6 months" type of content.
Native, especially on Windows and macOS, have been the domain of proprietary apps there's not much code outside of tutorials online to train a model on outside of official documentation.
I made a couple of small menu bar utilities for mac using Gemini, and it was OK at best. Kept wanting to use deprecated APIs, but with a lot of handholding I got them to work.
Would be neat to see Apple put out their own model specifically for Swift/SwiftUI
They say they're targeting Mac only for now, so it could be native code, or they could just have not tested/refined their prompt for other platforms yet.
> This was never the case for the desktop tools I mentioned.
I'd be curious how well Claude Code works for a native Swift app on macOS, if that's the platform you're on. I've found it extremely good at iOS apps so my guess is it would be equally good at building a native macOS app with the same stack.
I've tried using Codex and ChatGPT while working on a small SwiftUI app. It's not very good when it comes to newer APIs and features - I imagine due to lack of data about these things. Very often it would rather push something AppKit-based instead of SwiftUI.
It works, but feels really janky and messy.
I had one very annoying bug with file export API where extra view on export window would appear with a delay. No matter what I tried it didn't manage to fix it. Instead it would go on to try and completely rewrite whole file export class in various ways... which still didn't work as it claimed it would. Ended up fixing it manually by caching instance view locally.
> Also, as others mentioned, this just seems like Claude Code with extra steps, unless they managed to nail some sort of design standard enforcement they feel is better than what most people can get out of it.
Well yeah, isn’t that criticism we’ve had every LLM wrapper for years now? “Show me the prompt!” But that doesn’t mean these types of products are useless.
> If anything they do background processes better since you can send a very long task off to a server and shut off your computer, come back later and pick up where you left off.
I think it's fair to say that's a benefit of web apps over native apps in many cases. But for the kind of business app use case they're talking about, it's also a tradeoff. I can imagine a lot of business apps where you don't want to send the data to the server of a Replit etc. and doing all the processing local is a benefit.
> Also, as others mentioned, this just seems like Claude Code with extra steps, unless they managed to nail some sort of design standard enforcement they feel is better than what most people can get out of it.
My feeling is that it's intended for a less-technical audience than Claude Code.
If you're on Chrome and give them permission, or stuff them into Electron and friends, they can. The workflow isn't as smooth as with native applications, though.
On the other hand, the web browser does protect you from some of the risks this essentially "trust me bro" curl2bash-as-a-service product inherently comes with.
Might sound like a rube here, but: is agentic development really this good at novel UIs? The video shows a sort of cassette tape music player, and a fancy looking audio visualizer/equalizer thing. I'm well aware agents are very good at boilerplate UIs, but I wouldn't expect them to be able to one-shot novel, dynamic UI elements like this. I've had Claude attempt some SVG animations and the results were very crude. That was a year or so ago though. Are there established ways of letting agents iterate on UIs, i.e. having them visually verify the visual design and interactions?
I cannot bring myself to trust unreviewed software enough to install it on my own machine with arbitrary permissions. I understand the push for AI-generated websites, because the code running in my browser's sandbox is gonna have very limited permissions to do anything evil, but desktop apps are a completely different story
I feel you! We thought about this and all apps will have a permission model. So you can limit it to specific file disk locations, domains for network requests, and so on.
I don't see how this solves the issue, something bad can happen regardless of permission granularity, no?
Definitely a good initiative though. I like how coding harnesses do it, showing you the exact command that would run, or running it in a sandbox first.
Not to be a curmudgeon, but why are they spending time on this? As an enthusiastic Raycast user, I would prefer to see them focus on making Raycast better, not finding new ways to jump on the AI bandwagon.
Welp this is what happens when the USA is spending more into this than it did in the space race comparatively. Space race we got to the moon, the AI output has yet to show profit from businesses other than funding the input.
I had the same reaction. They've had Raycast releases paused for some time to focus on large feature improvements, but I wondered if it was partly for this.
Claude Code is pretty good at Swift + Swift UI. I created and have been iterating on a menubar app for myself that I plan to share with a small team. I'd prefer to do this native than go through a 3rd party solution.
I do have prior experience developing for iOS but that was pre-swift.
I'm also just working on my first iOS Swift app (Mostly for myself, don't know yet if I'll make it public as it's just a clone of Swarm / Gowalla but based on OpenStreetMap data) and it works really well with Claude Code.
I'm not using the Xcode integration and so there's still some rough parts where build errors show up in Xcode and I then have to paste them into my Terminal.
When you are used to backend work...it's kinda fun to see an app come to life and run on your phone though.
xcode's new AI using claude is not performing as well as claude code for me. I've tried a couple times and quickly fall back to using vscode with xcode sitting in a window beside. I don't mind the copypaste of warnings and errors since my workflow is less vibe and more directed/iterative.
Someone please exfiltrate their prompts/skills so i can use these on Codex. I've have relative success building my own apps for mac using Codex but they're uglier than sin and dont seem to understand well how permissions work.
Having antigravity is useful because Gemini 3.1 is pretty good at generating UI sugar. Claude 4.6 Opus provides nothing to write home about. Their shadcn looksmaxxing hasn't generalized to writing good desktop UIs.
Raycast's only edge here seems to be the fact that they are obviously very good at Mac app development and probably have impeccable skills/documentation for building them.
Taking a step back, it's pretty clear that Raycast is angling for an Apple acquisition here with this play. If I'm Apple, the reason to buy a product/team like this is a no-brainer.
> Taking a step back, it's pretty clear that Raycast is angling for an Apple acquisition here with this play.
No kidding, although I think Apple would only be interested if it uses SwiftUI. (The marketing page doesn't say. Raycast itself uses React + Node for extensions, but its React components render to native widgets.)
A similar product in the mobile space is Rork - I haven't used it but I've seen it on twitter a bit. I definitely wouldn't be surprised to see Apple acquire one of them.
But can it manage external libraries or use only the existing sdk? I had a non tech friend run into an issue recently where she wanted to automate a pdf action. Eventually I realized she needed to run homebrew and install a library. Curious if this actually manages that kind of process.
Just what absolutely no one needed: another locked down and non web platform with horrific security that tries to digitally enslave people just the tiniest level above what they can accept now. I don’t see any future where raycast can survive and i would say its a good thing.
I understand some of the skepticism towards this product, but are you saying this will somehow negatively impact Raycast (the company)? Raycast the tool is incredibly useful, so I'm surprised to see this sentiment.
Looking forward to trying this out and see how this differs from more manual approaches. One thing that stands out is an included store for public/private distribution — that’s super convenient given the cumbersome (and maybe, horrific) process that is the app store submission.
So it looks like they’re creating their own App Store within the app? At least it’s kept separate from official apps. But also how is that not a security nightmare Apple won’t allow?
One would think it must otherwise there are all these issues with compiling, signing tc if they don’t have xcode installed etc. I would guess it’s some webview wrapper with a layer to expose desktop app functionality
Impressive feat. Definitely not for me though, and for sure I won't be there to debug one of these when my parents will call me because it broke their computers.
A local-only voice to text whisper.cpp transcriber I can globally use while holding ctrl-semicolon.
A menubar app that manages blocky and can easily turn it off or change dns.
A tool like hammerspoon but I configure it via nix-darwin and it has no cruft.
All of these are apps that use 30MB memory and are better than the apps they replace, and I can make changes any time I want. That's far better than using someone else's software and giving it privileged access to my machine.
Also, perhaps the best point is that so much software is junk that is obsoleted by someone with better UX intuitions even if they are vibe-coding it. Being written by hand by an engineer means basically nothing when it comes to "is this a good app?" Which is why product-minded people are the biggest winners in the new AI era.
The problem that software suffers from is that every app/program tries to cover as many bases and use cases as possible in a single package. Obviously it's what you want to do if you want to maximize reach/customers.
Vibe apps are different. They do exactly what you want, exactly the way you want it done. No more downloading an app that is mysteriously 180MB and requires watching a youtube video to learn how to make it change your background every 5 minutes to different dog pictures.
I can think of at least 1 major improvement to so many of the apps I use day to day.
Desktop software is nowhere near good enough to consider random usecases "already done". Not that glaze looks particularly special, but there's so many improvements the desktop experience begs for.
An easy to use cross-platform GUI builder for one. Even something as basic as a calendar app doesn't have a clear obvious winner today.
So, another wrapper around claude 4.6 for +xx% higher price? Using just claude code, one can do what glazeapp seems to aim for, no? "Beautiful by default" seems to be a system prompt akin to:
Design Philosophy Create apps that feel premium, polished, and worthy of being featured on Dribbble's most popular shots. Every pixel matters. White space is your friend. Less is more, but what remains must be perfect.
Visual Design Principles
Color & Theming
• Use sophisticated, limited color palettes (2-3 primary colors maximum) ...
It looks like it's a lot of sensible defaults UI libraries to use, UX framework presets, etc, designed for an end user who doesn't know what Node or Electron or Rust or Tauri are. Plus, the page describes an app sharing mechanism as well built-in.
To be honest, but I would love to have some ecosystem around building apps which lets me share my custom apps with team members in my organization. Without having to take care of updating, provisioning, and distributing the app, etc.
even better if the apps are not electron bloated and mac-native.
Reading what they're offering, the stand-out to me is making publishing the applications easy for others on your team to use. That would he a pain point for non technical users.
I get your point. But if sharing with others is a vital part of this, then ... they'd be better off sticking to web apps instead :)
"Create with glaze, hit publish and we'll give you an url".
Even though they portray some of the benefits of this app as unique to desktop apps, they're not (e.g. storing data on device, offline mode etc.).
Am not a hater. Love Raycast. Saw the post and opened the link intrigued what they came up with and was somewhat disappointed tbh. Good luck to them anyway!
I think their value add if you’re comfortable with Claude Code is probably some nice tooling for the packaging, and they probably sign apps for you too?
1. macOS and Windows require installation of Xcode and Visual Studio respectively, and if in Apple's case you kinda can install these tools headlessly and choose to install only the "build tools" package, Microsoft's creature is gonna daze and confuse you with a crap-ton of checkboxes and no easy "just install whatever is minimally needed to compile my code" button, and I don't recall if there is way to install build tools on Windows through terminal.
2. what is going to be distributed? source code itself or actual binaries? and what will the security model of Glaze store be? same as extensions, "everything is open-source and undergoes Raycast's and community review"?
3. Glaze is going to come to Windows and Linux, if we trust the Q&A section at the end. what will Glaze build upon? separate frameworks and languages for each platform or something multi-platform [1] like Tauri or Kotlin Multiplatform? or are you going to copy the Raycast extension model - just run Node, expose some platform integration, and parse React render trees through "Glaze Runtime"? I've been working on a bug in Vicinae [2][3], and I've seen this model in action. it's very hard to make it perform well, but all it takes to achieve native look and feel is to just map React render trees to whatever system component OS offers. (in Vicinae's case, it's Qt. bet that it's done with SwiftUI on macOS and WinUI 3 on Windows.)
[1]: there is a difference between "cross-platform" and "multi-platform". "cross-platform" means "I behave equally across platforms and have no awareness of native look and feel" (e.g. Electron, Unity, Flutter), while "multi-platform" means "I can adapt across platforms to the degree you need" (e.g. C/C++, Rust, KMP)
As an interesting counter-proposal to wasting time with this... look for older less popular/downloaded/featureful apps written by people for their own education, edification and enjoyment.
They may not work the way you wish they would, but you can learn a lot from them, be inspired by them, and leave feedback.
That's how you actually encourage more people to get started and continue making their own tools.
I love Raycast. I would probably be called a "power user" - I use it all day long and have a fairly sophisticated and customized configuration and set of workflows. Raycast is actually one of the primary things keeping me on MacOS these days (please release a Linux version!).
I am worried this is the start of them trying to diversify their product offering because revenue has stalled in the core Raycast product and VC demands more returns. I don't want to be jaded, but history teaches me to be. Here's hoping that Raycast itself is still a focus for the company.
I thought this must be a joke at first. "Glaze" is in pretty heavy use as recent slang for "when someone excessively praises another person in a way that feels over-the-top." https://creativesimiles.com/glaze-meaning-slang/
It's so much worse. Your link fails to mention that the "Glaze" in question is a cough bodily fluid. Yes that one. Have I seen politicians use "glaze" recently? Yes. Gross.
On the other hand it is kind of the perfect name for Yet Another AI Website Maker (YAAWM?).
While I don’t think the new meaning is incredibly widespread yet, it’s not uncommon for words to change meaning over time. I wouldn’t be surprised if a decade or two from now, the original meaning has been mostly forgotten.
> First you said all you want is love and affection / Let me be your angel and I'll be your protection / … / Thought I was a donut, you tried to glaze me
> I ain’t gotta tell you I had a Dunkin' Donuts fetish back in the day. I used to get a dozen donuts every day, man. So it was one of the things that was on my mind
I don't understand how these type of projects are still tried and get any traction... anyone who has tried them will 100% know it won't go beyond a happy path demo. If they want to seriously use/publish the app beyond playing around, it'll require weeks of iteration via AI, which will cost you an arm and a leg in tokens.
Founder of Raycast here so obviously biased but you’ll be surprised. You get a working app one-shotted pretty much all the time. Sure if it is something more complex you might need a few more prompts. Just to give you some examples on what we’ve seen: - Our support team runs on Glaze apps to review Raycast extensions. It connects to GitHub, checks out code locally, gets realtime updates and so on. - The sound agency build a functioning synthesiser for the launch video. It works even with MIDI. - We’re about to cancel a team-wide subscription and replace it with a Glaze app.
Not everything is possible yet and sure more complex things need more prompts but you’ll be surprised what Glaze is capable of already. It’s day one…
> You get a working app one-shotted pretty much all the time.
Can you one-shot a raycast alternative with this? This'll be the real test.
So… could I one-shot a Glaze competitor? ;)
More seriously, what do you believe your moat is here?
There will probably be a few of these like TextEditors. I already built this and have features in mind that I’m not sure Glaze is thinking of.
This sounds promising. If I may take advantage of you being here, what language does it write in? Does it build genuine native apps (Cocoa, WinAPI or WinUI, etc) or Electron?
The FAQ was light on technical details. But I am someone keen to read all the technical details :)
Does it generate native apps, or just Electron?
I haven't used v0 or replit before, I have the same feelings as you. But I've been thinking about building macOS apps for my personal use for a long time now. Also I'm a long time Raycast user. I have a bias here, so I've joined the waitlist, I can't be sure until I try, right?
It's certainly a nice promotional website.
My first thought was, "So, Replit and ilk?", seems they expected that comparison:
> How is Glaze different from Lovable, Replit, or v0?
> Those tools build for the browser. Glaze builds for your desktop. That means your apps can access your file system, your camera, keyboard shortcuts, menu bar integration, and background processes. Things a web app can’t do. It’s a different category entirely.
Pretty sure modern web apps can do all of those (sans menu bar). (If anything they do background processes better since you can send a very long task off to a server and shut off your computer, come back later and pick up where you left off.)
Also, as others mentioned, this just seems like Claude Code with extra steps, unless they managed to nail some sort of design standard enforcement they feel is better than what most people can get out of it.
The quick publishing is kind of nice, but it immediately made me think it would be more interesting to have a way to quickly remix other people's creations, similar to the Figma Community tab: you can take someone else's work, break it apart to see how it works, then tweak it how you want it.
I took a few shots at building desktop apps with Tauri, Wails and Electron using Claude Code, and the results were not very good at all. In fact, they were by far the worst results I've gotten with the tool. I can easily clone one of my boilerplate repos in Rails, or Django and prompt away, and the results are consistently good, as in, functional MVP in a few hours. This was never the case for the desktop tools I mentioned.
This looks like a highly specialized tool for desktop that actually works. I watch the demo and I am assuming the apps are actually made with some kind of technology a la Tauri, or Electron, thus making the apps cross-platform.
I don't think we are anywhere near a tool like this for native, but that's a lost battle anyway.
I’ve had a totally different experience. I’ve coded 3 different Tauri apps and 1 Wails app with Claude Code and it was some of the easiest work I’ve done with AI assisted coding. That said, the local features that Rust is handling in the Tauri app is not anything heavy, just moving files around, some regex matching, and some SQLite stuff. All of the headache I had in these apps was the React frontends and Node issues. The Rust features all worked pretty much first try every time.
> I don't think we are anywhere near a tool like this for native, but that's a lost battle anyway.
I hope it's not a lost battle, tbh. I was hoping with AI & Vibe Coding we'd see sort of a resurgence of native first desktop apps, but so far it's just all been a continuation of the web app & web tech hegemony.
Maybe not for Windows as their native GUI story is a lost cause now, but for sure macOS and I had hopes of it leading to a renaissance of desktop linux apps in GTK instead of electron, but that (the Linux) community seems to be hostile to any AI generated code at all for now.
I have been seeing more and more native desktop apps in the past few months (octarine for instance), but most of them would've honestly been better off as web-apps, or at least a polished electron app.
> seems to be hostile to any AI generated code at all for now.
Because the majority of vibe-coded apps are low effort.
Well, to be fair, I do have an experience working on a Windows Forms app from scratch. App connects to a very specific scanner via customs drivers and makes use of a remote API for data tasks. The app works, it's stable, but I'm not going to lie, AI assisted coding for this particular stack does require a very large amount of nurturing, it is just not the same experience you get with web apps. Nevertheless, it did it.
Makes sense. There's plenty of freely available code and data online for using web tech. Any number of free online bootcamps spawned in the mid 2010s are full of "Become a React developer in 6 months" type of content.
Native, especially on Windows and macOS, have been the domain of proprietary apps there's not much code outside of tutorials online to train a model on outside of official documentation.
I made a couple of small menu bar utilities for mac using Gemini, and it was OK at best. Kept wanting to use deprecated APIs, but with a lot of handholding I got them to work.
Would be neat to see Apple put out their own model specifically for Swift/SwiftUI
They say they're targeting Mac only for now, so it could be native code, or they could just have not tested/refined their prompt for other platforms yet.
> This was never the case for the desktop tools I mentioned.
I'd be curious how well Claude Code works for a native Swift app on macOS, if that's the platform you're on. I've found it extremely good at iOS apps so my guess is it would be equally good at building a native macOS app with the same stack.
I've tried using Codex and ChatGPT while working on a small SwiftUI app. It's not very good when it comes to newer APIs and features - I imagine due to lack of data about these things. Very often it would rather push something AppKit-based instead of SwiftUI.
It works, but feels really janky and messy.
I had one very annoying bug with file export API where extra view on export window would appear with a delay. No matter what I tried it didn't manage to fix it. Instead it would go on to try and completely rewrite whole file export class in various ways... which still didn't work as it claimed it would. Ended up fixing it manually by caching instance view locally.
Why not use SwiftUI or whatever is native to the platform?
> Also, as others mentioned, this just seems like Claude Code with extra steps, unless they managed to nail some sort of design standard enforcement they feel is better than what most people can get out of it.
Well yeah, isn’t that criticism we’ve had every LLM wrapper for years now? “Show me the prompt!” But that doesn’t mean these types of products are useless.
> If anything they do background processes better since you can send a very long task off to a server and shut off your computer, come back later and pick up where you left off.
I think it's fair to say that's a benefit of web apps over native apps in many cases. But for the kind of business app use case they're talking about, it's also a tradeoff. I can imagine a lot of business apps where you don't want to send the data to the server of a Replit etc. and doing all the processing local is a benefit.
You mean “fork” other apps.
A big thing would be API requests/browser automation. Web apps can’t do that without a backend proxy due to CORS
> Also, as others mentioned, this just seems like Claude Code with extra steps, unless they managed to nail some sort of design standard enforcement they feel is better than what most people can get out of it.
My feeling is that it's intended for a less-technical audience than Claude Code.
I can certainly see that. If they really did manage to make some really effective design tooling, would be a great candidate for an MCP server.
> Pretty sure modern web apps can do all of those
If you're on Chrome and give them permission, or stuff them into Electron and friends, they can. The workflow isn't as smooth as with native applications, though.
On the other hand, the web browser does protect you from some of the risks this essentially "trust me bro" curl2bash-as-a-service product inherently comes with.
Ironically, there's another project named Glaze, that aims to "protect artists from generative AI" (https://glaze.cs.uchicago.edu/)
There's also a window(s) tiling manager named Glaze that's pretty popular: https://github.com/glzr-io/glazewm
Might sound like a rube here, but: is agentic development really this good at novel UIs? The video shows a sort of cassette tape music player, and a fancy looking audio visualizer/equalizer thing. I'm well aware agents are very good at boilerplate UIs, but I wouldn't expect them to be able to one-shot novel, dynamic UI elements like this. I've had Claude attempt some SVG animations and the results were very crude. That was a year or so ago though. Are there established ways of letting agents iterate on UIs, i.e. having them visually verify the visual design and interactions?
I cannot bring myself to trust unreviewed software enough to install it on my own machine with arbitrary permissions. I understand the push for AI-generated websites, because the code running in my browser's sandbox is gonna have very limited permissions to do anything evil, but desktop apps are a completely different story
I feel you! We thought about this and all apps will have a permission model. So you can limit it to specific file disk locations, domains for network requests, and so on.
I don't see how this solves the issue, something bad can happen regardless of permission granularity, no?
Definitely a good initiative though. I like how coding harnesses do it, showing you the exact command that would run, or running it in a sandbox first.
Can you explain how the permission model works?
Not to be a curmudgeon, but why are they spending time on this? As an enthusiastic Raycast user, I would prefer to see them focus on making Raycast better, not finding new ways to jump on the AI bandwagon.
Welp this is what happens when the USA is spending more into this than it did in the space race comparatively. Space race we got to the moon, the AI output has yet to show profit from businesses other than funding the input.
I had the same reaction. They've had Raycast releases paused for some time to focus on large feature improvements, but I wondered if it was partly for this.
the first scenario that came to mind is that they built it for themselves and then open sourced it
Claude Code is pretty good at Swift + Swift UI. I created and have been iterating on a menubar app for myself that I plan to share with a small team. I'd prefer to do this native than go through a 3rd party solution.
I do have prior experience developing for iOS but that was pre-swift.
I'm also just working on my first iOS Swift app (Mostly for myself, don't know yet if I'll make it public as it's just a clone of Swarm / Gowalla but based on OpenStreetMap data) and it works really well with Claude Code.
I'm not using the Xcode integration and so there's still some rough parts where build errors show up in Xcode and I then have to paste them into my Terminal.
When you are used to backend work...it's kinda fun to see an app come to life and run on your phone though.
xcode's new AI using claude is not performing as well as claude code for me. I've tried a couple times and quickly fall back to using vscode with xcode sitting in a window beside. I don't mind the copypaste of warnings and errors since my workflow is less vibe and more directed/iterative.
Same thought I had while reading, don’t really see a big advantage here.
Thank you for your service.
We need more of you. Not more electron slop.
Someone please exfiltrate their prompts/skills so i can use these on Codex. I've have relative success building my own apps for mac using Codex but they're uglier than sin and dont seem to understand well how permissions work.
Having antigravity is useful because Gemini 3.1 is pretty good at generating UI sugar. Claude 4.6 Opus provides nothing to write home about. Their shadcn looksmaxxing hasn't generalized to writing good desktop UIs.
Raycast's only edge here seems to be the fact that they are obviously very good at Mac app development and probably have impeccable skills/documentation for building them.
Taking a step back, it's pretty clear that Raycast is angling for an Apple acquisition here with this play. If I'm Apple, the reason to buy a product/team like this is a no-brainer.
Raycast are not building Mac apps the apple way though. They are using react native and I am willing to bet that this does too.
>Taking a step back, it's pretty clear that Raycast is angling for an Apple acquisition here with this play.
Raycast recently made a Windows version. So perhaps they aren't as Apple-centric.
> Taking a step back, it's pretty clear that Raycast is angling for an Apple acquisition here with this play.
No kidding, although I think Apple would only be interested if it uses SwiftUI. (The marketing page doesn't say. Raycast itself uses React + Node for extensions, but its React components render to native widgets.)
A similar product in the mobile space is Rork - I haven't used it but I've seen it on twitter a bit. I definitely wouldn't be surprised to see Apple acquire one of them.
But can it manage external libraries or use only the existing sdk? I had a non tech friend run into an issue recently where she wanted to automate a pdf action. Eventually I realized she needed to run homebrew and install a library. Curious if this actually manages that kind of process.
My metric for this kind of stuff is: Did Glaze build the Glaze app?
Just what absolutely no one needed: another locked down and non web platform with horrific security that tries to digitally enslave people just the tiniest level above what they can accept now. I don’t see any future where raycast can survive and i would say its a good thing.
I understand some of the skepticism towards this product, but are you saying this will somehow negatively impact Raycast (the company)? Raycast the tool is incredibly useful, so I'm surprised to see this sentiment.
Looking forward to trying this out and see how this differs from more manual approaches. One thing that stands out is an included store for public/private distribution — that’s super convenient given the cumbersome (and maybe, horrific) process that is the app store submission.
Getting closer to ClaudeVM https://jperla.com/blog/claude-electron-not-claudevm
So it looks like they’re creating their own App Store within the app? At least it’s kept separate from official apps. But also how is that not a security nightmare Apple won’t allow?
This looks super fun, actually.
I wonder what it is actually building. Tauri apps, maybe?
One would think it must otherwise there are all these issues with compiling, signing tc if they don’t have xcode installed etc. I would guess it’s some webview wrapper with a layer to expose desktop app functionality
Or it’s compiled in the cloud?
Electron I bet
It is interesting how so many different companies end up converging to some sort of AI coding.
Raycast -> Glaze AirTable -> Lovable Competitor Retool -> Lovable Competitor
Even those early in the journey are converging towards coding.
Have had good results on MacOS just using codex (or your cli of choice).
Have it create a swift app, unless extended permissions are needed it can compile withouy going into xcode.
Few simple util apps, disk cleaner, clipboard manager. Worked pretty well.
Had better results than using xcode's built in ai extension.
There seem to be more AI app building platforms than actual apps being built these days.
No mention of security.. remarkable
The “S” in “Glaze” stands for “security”.
Makes sense to me. It's a marketing page. Know your audience.
They'd need another 30 full time devs for that.
I can't imagine trusting these apps with access to my camera, file system or any other sensitive permissions.
Impressive feat. Definitely not for me though, and for sure I won't be there to debug one of these when my parents will call me because it broke their computers.
This is just a landing page. There's not even any decent product specs. Nothing technical. How does this make front page of hacker news?
It will be awesome if these were native apps instead of JavaScript apps. It's not mentioned anywhere explicitly that these are native.
How many apps do you really need that are not already done - perhaps even better?
I've vibe coded all sorts of apps for my macbook.
A better replacement to iStat Menus.
A local-only voice to text whisper.cpp transcriber I can globally use while holding ctrl-semicolon.
A menubar app that manages blocky and can easily turn it off or change dns.
A tool like hammerspoon but I configure it via nix-darwin and it has no cruft.
All of these are apps that use 30MB memory and are better than the apps they replace, and I can make changes any time I want. That's far better than using someone else's software and giving it privileged access to my machine.
Also, perhaps the best point is that so much software is junk that is obsoleted by someone with better UX intuitions even if they are vibe-coding it. Being written by hand by an engineer means basically nothing when it comes to "is this a good app?" Which is why product-minded people are the biggest winners in the new AI era.
Neat! What does the stack look like?
The problem that software suffers from is that every app/program tries to cover as many bases and use cases as possible in a single package. Obviously it's what you want to do if you want to maximize reach/customers.
Vibe apps are different. They do exactly what you want, exactly the way you want it done. No more downloading an app that is mysteriously 180MB and requires watching a youtube video to learn how to make it change your background every 5 minutes to different dog pictures.
I can think of at least 1 major improvement to so many of the apps I use day to day.
Desktop software is nowhere near good enough to consider random usecases "already done". Not that glaze looks particularly special, but there's so many improvements the desktop experience begs for.
An easy to use cross-platform GUI builder for one. Even something as basic as a calendar app doesn't have a clear obvious winner today.
Literally hundreds.
In the DOS days, I would have whipped them up in BASIC. This was standard practice for PC users who were not "software engineers" by trade.
The complication of PCs over the past 30+ years have robbed regular users of this ability.
Tools like this close the gap, and that's awesome.
how many problems do you have unsolved?
So, another wrapper around claude 4.6 for +xx% higher price? Using just claude code, one can do what glazeapp seems to aim for, no? "Beautiful by default" seems to be a system prompt akin to:
It looks like it's a lot of sensible defaults UI libraries to use, UX framework presets, etc, designed for an end user who doesn't know what Node or Electron or Rust or Tauri are. Plus, the page describes an app sharing mechanism as well built-in.
To be honest, but I would love to have some ecosystem around building apps which lets me share my custom apps with team members in my organization. Without having to take care of updating, provisioning, and distributing the app, etc.
even better if the apps are not electron bloated and mac-native.
Reading what they're offering, the stand-out to me is making publishing the applications easy for others on your team to use. That would he a pain point for non technical users.
I get your point. But if sharing with others is a vital part of this, then ... they'd be better off sticking to web apps instead :) "Create with glaze, hit publish and we'll give you an url".
Even though they portray some of the benefits of this app as unique to desktop apps, they're not (e.g. storing data on device, offline mode etc.).
Am not a hater. Love Raycast. Saw the post and opened the link intrigued what they came up with and was somewhat disappointed tbh. Good luck to them anyway!
I think their value add if you’re comfortable with Claude Code is probably some nice tooling for the packaging, and they probably sign apps for you too?
so many unknowns...
1. macOS and Windows require installation of Xcode and Visual Studio respectively, and if in Apple's case you kinda can install these tools headlessly and choose to install only the "build tools" package, Microsoft's creature is gonna daze and confuse you with a crap-ton of checkboxes and no easy "just install whatever is minimally needed to compile my code" button, and I don't recall if there is way to install build tools on Windows through terminal.
2. what is going to be distributed? source code itself or actual binaries? and what will the security model of Glaze store be? same as extensions, "everything is open-source and undergoes Raycast's and community review"?
3. Glaze is going to come to Windows and Linux, if we trust the Q&A section at the end. what will Glaze build upon? separate frameworks and languages for each platform or something multi-platform [1] like Tauri or Kotlin Multiplatform? or are you going to copy the Raycast extension model - just run Node, expose some platform integration, and parse React render trees through "Glaze Runtime"? I've been working on a bug in Vicinae [2][3], and I've seen this model in action. it's very hard to make it perform well, but all it takes to achieve native look and feel is to just map React render trees to whatever system component OS offers. (in Vicinae's case, it's Qt. bet that it's done with SwiftUI on macOS and WinUI 3 on Windows.)
[1]: there is a difference between "cross-platform" and "multi-platform". "cross-platform" means "I behave equally across platforms and have no awareness of native look and feel" (e.g. Electron, Unity, Flutter), while "multi-platform" means "I can adapt across platforms to the degree you need" (e.g. C/C++, Rust, KMP)
[2]: https://github.com/vicinaehq/vicinae
[3]: https://github.com/vicinaehq/vicinae/pull/1158
what about barebone/starter desktop app that can be modified itself by prompts?
that's would be Electron app, but without unneeded bloat
Vibe-coding desktop apps is a much, much better solution for the vast majority of one-off tools most users want to build.
No thanks.
As an interesting counter-proposal to wasting time with this... look for older less popular/downloaded/featureful apps written by people for their own education, edification and enjoyment.
They may not work the way you wish they would, but you can learn a lot from them, be inspired by them, and leave feedback.
That's how you actually encourage more people to get started and continue making their own tools.
they did it again, glad I am on Mac, congrats raycast
I love Raycast. I would probably be called a "power user" - I use it all day long and have a fairly sophisticated and customized configuration and set of workflows. Raycast is actually one of the primary things keeping me on MacOS these days (please release a Linux version!).
I am worried this is the start of them trying to diversify their product offering because revenue has stalled in the core Raycast product and VC demands more returns. I don't want to be jaded, but history teaches me to be. Here's hoping that Raycast itself is still a focus for the company.
Is this just shitting out electron crap?
"Insecure apps, reimagined by you"
I thought this must be a joke at first. "Glaze" is in pretty heavy use as recent slang for "when someone excessively praises another person in a way that feels over-the-top." https://creativesimiles.com/glaze-meaning-slang/
ie the annoying way that LLMs interact with users
It's so much worse. Your link fails to mention that the "Glaze" in question is a cough bodily fluid. Yes that one. Have I seen politicians use "glaze" recently? Yes. Gross.
On the other hand it is kind of the perfect name for Yet Another AI Website Maker (YAAWM?).
While I don’t think the new meaning is incredibly widespread yet, it’s not uncommon for words to change meaning over time. I wouldn’t be surprised if a decade or two from now, the original meaning has been mostly forgotten.
It's worse than that, the glaze is only the by-product of the primary action in question.
Nah it’s from Dunkin Donuts [1].
> First you said all you want is love and affection / Let me be your angel and I'll be your protection / … / Thought I was a donut, you tried to glaze me
> I ain’t gotta tell you I had a Dunkin' Donuts fetish back in the day. I used to get a dozen donuts every day, man. So it was one of the things that was on my mind
[1]: https://genius.com/1716467
> https://creativesimiles.com/glaze-meaning-slang/
Kids these days are always saying "snaoƨd" and "foʀaṅr"
> Written By Lucas Gray
There is no way a human wrote that page. If Lucas Gray even exists, he should probably reconsider that last image, and his life more generally.
"Heavy use" to the point where it might even be more common than the older senses of "cover with icing" or "install windows".
Related, "glass" or "glassing" can also refer to stabbing someone in the face with broken glass or decimating a world in nuclear holocaust.
Maybe a quietly dissenting PM snuck it by. If so, nice.