I'm generally a big fan of zed and have been using it for 60%ish of my dev time for 6 months or so. A couple of nice things to note:
- It really is remarkably responsive,and makes one really notice how UNresponsive everything else is. I have reasonably fast machines, so we're not talking about the difference between 5ms typing lag and 500ms, but it's still pretty surprising. VSCode never felt slow on my macs until I started using Zed.
- They seem reasonably responsive to feedback. There was some contention around how search/replace was initially implemented, and the current builds have something much more usable IMO. I'm not sure how much that was driven by community feedback, but the changes were great.
- The debug syntax tree mode is a really neat feature that I think demonstrates how much more advanced zed is under the hood than older editors that are doing syntax highlighting via regex.
There are a few downsides that I'm hoping get addressed soon:
- The collaboration workflow/security isn't very clear to me. You sign in via github (no other option???), there are 'contacts' (I guess these are github usernames?), and 'channels' (where do these live? on zed's servers?). I would really like to know if I can self-host the chat server and use a company oauth provider rather than github. If the diffs being passed around are going through zed's servers, that may be a showstopper for the company I work for as well. If they're p2p and encrypted, maybe not.
- I would love to see ollama integration. This + continue is the only reason why I spend any amount of time in vscode now. There's an issue for it here: https://github.com/zed-industries/zed/issues/4424
I’m really picky about my tools in general and latency in particular, and I give Zed a spin from time to time, it’s sweet. Clean, minimal design aesthetic, tree-sitter, tight code, it’s a nice bit of work and I dig it a lot.
But emacs 29 with the right flags and a tuned GC (no one does this! it’s got a heap-size from the 80s!) is just as snappy and has more amazing packages than VSCode.
There’s a market for people who want something snappier than VSCode but less labor-intensive to set up than emacs, and I wish them luck: I think it’s a front runner there. But I can’t imagine switching my main axe up with a holy shit moment a lot crazier than tree-sitter in 2024 and not having the render loop be in JS.
People are asking what settings would be useful to tweak. Check out the following:
- gcmh -package and gcmh-mode and/or gc-cons-threshold variable (former should take over the latter)
- read-process-output-max
- jit-lock-defer-time
- package-native-compile
Doom Emacs sets gcmh in its initialization so tweaking that might not be needed there. You may still want to touch gcmh-high-cons-threshold and gcmh-idle-delay-factor. Here are mine currently:
I've done nothing scientific to check out if these help at all, though, so take them with salt. With and without these, emacs seems quite sluggish at least on a Macbook, in certain modes. On Linux things seem to be a bit better.
The thing about gc-cons-threshold is that, too low a value and it will collect garbage too frequently. While a high value will drastically reduce frequency, but then GC pause will be big and you will likely feel whole system freeze and stutter when it happens. For me, 1GB is too high because I sorta notice the pause, I set it to 32MB but completely disable GC when minibuffer is active:
Anyway, I also have these in my init (I nicked these from Doom which I don't use but they have done a lot of work to dive deep into these things, mostly around what improves start up but the following might help in general):
;; if you don't use RTL ever, this could improve perf
(setq-default bidi-display-reordering 'left-to-right
bidi-paragraph-direction 'left-to-right
bidi-inhibit-bpa t)
;; improves terminal emulator (vterm/eat) throughput
(setq read-process-output-max (* 2 1024 1024)
process-adaptive-read-buffering nil)
(setq fast-but-imprecise-scrolling t
redisplay-skip-fontification-on-input t
inhibit-compacting-font-caches t)
(setq idle-update-delay 1.0)
How do you get a reasonably fancy Emacs to start up/open files quickly? Every time I try to get into EMacs after adding a few packages it becomes painfully slow to open a file for editing. Then I try to understand Emacs server, then I fail / give up and go back to vim.
I've been using emacs as my main editor for years now and don't find the startup speed to be an issue - to the point I've never bothered with emacsclient:
Emacs server is quite straightforward. M-x server-start in a running emacs, then emacsclient <filename>. emacsclient -c if you want to open the file in a new window, or emacsclient -t if you want to open it in your current terminal. C-x # when you're done editing the file.
The way the other guy said. Then it breaks enough times, you give up, and switch to Helix. Nothing snappier out there (vanilla Vim included) and it comes with all the goodies you'd install 100 packages for (DAP, LSP, treesitter, sane default configs for dozens of languages)
To each their own, but for a hardcore Emacs user (wrote a number od my own plugins) that actually tried and couldn't seriously adopt VSCode for more than a few months, Helix was such a nice surprise + change of pace. Worth giving it a shot if you're caught in that gap.
Is there something that approximates org-mode outlining? I don't need all the other stuff org-mode does, but just the simple outliner functionality combined with all the other Emacs stuff is my golden handcuffs.
Not that I'm aware of just yet, unfortunately, so might be a good reason to stick around. For me the switch coincided with jumping on the zettelkasten train and I haven't looked back since. I miss some org mode fratures but constantly `org-refile`-ing I can do without.
1.4 seconds for 100 packages, half of which are on probation? My (and soon `HYPER//MODERN`'s) package set is under heavy construction, I'll be sad if I don't get it under 300ms including loading and rendering a logo at 6k from a cold disk which that was.
Rust is a perfectly good language for writing tight code, but those `emacs` inner loops have been tuned by hard-ass pros for 30+ years in straight C, which is plenty fast too. Even the best Rust code doesn't have that kind of tuning in.
With no intention to push back on Emacs being highly optimized, age of an application is not an accurate proxy for level of performance. If it were, grep would be faster than ripgrep. New software can apply optimisations discovered over the previous 40 years, and leverage architectural designs with different assumptions, like the GC limits in Emacs discussed elsewhere in the comments. Older applications can apply those same optimizations, but it's not a given that they will, and changing their architecture can be daunting, for sometimes unknown benefits.
I was about to answer "Use emacsclient", but then I realized you can't be bothered to configure it. For me, putting "(server-start)" into .emacs was always enough, but YMMV.
It is a great sentence, and that holy shit moment is coming: when JS gains all the expressivity of tree-sitter and uses it to render code more beautifully than it has ever been rendered before.
The, uh, GNU folks, are a little conservative in some ways that surprise the mainstream developers. But there are always flags for it. Well, except stupid-ass dynamic-linking of `glibc`.
> are a little conservative in some ways that surprise the mainstream developers
Conservative is one thing, but having the default be something that's detrimental to 99.999% of your users in 99.95% of situations feels like bad defaults/design, not being conservative. My worthless 2 eurocents :-)
Both @darthrupert and @natyrys on the thread have said most of what I know. It's a little machine dependent to get the true sweet spot, but raising it from the like, megabyte (slight hyperbole) or whatever that it defaults to is the 80/20.
(setq gc-cons-threshold (\* 100 1024 1024) ;; GC sometime after allocating 100Mb
read-process-output-max (\* 1024 1024)
company-idle-delay 0.0 ;; company completions should be fast
company-minimum-prefix-length 1
lsp-idle-delay 0.1) ;; clangd is fast
(run-with-idle-timer 2 t (lambda () (garbage-collect))) ;; Trigger a GC after 5s of idle time
This is something of a hack of course, but it works & memory is cheap. You might want to push the GC threshold back down after emacs startup - a GC that fires early is a GC that does less work per invocation & is therefore less likely to cause perceptible stutter.
i got annoyed by emacs gc pauses and just disabled the gc completely (i think). haven't noticed any downsides so far. i'm in emacs all day...
;; don't garbage collect based on cons count
(setq gc-cons-threshold 10000000000)
(defun garbage-collect (&rest args)
(message "trying to garbage collect. probably you want to quit emacs."))
something i love about emacs is everything is implemented in the same abstraction (of a "buffer of text"). so i can manipulate and move around an embedded terminal in the same way as a normal file, etc.
playing briefly with Zed it seemed like the terminal was a totally different "thing" than a normal file; i couldn't run the same text selection operations, couldn't split the window vertically or horizontally, etc. to me that loses a ton of the benefit of having a terminal in the editor in the first place.
still, it's a very interesting project. i have a very love/hate relationship with emacs, so i'm always interested in alternatives...
I mentioned it above but I’m flattered by your interest (and remember your username as someone serious): my old dots were ok and are floating around but we’re putting together a multi-editor curated setup for good defaults and an emphasis on good, available-weight models as first-class citizens.
It’s a 404 until I get to an RC, but aiming for March 1st on “hyper-modern.ai” for all MIT on emacs, nvim, and vscode.
Most examples you'll see of `straight.el`/`use-package` configurations will have some version of: `(use-package something-cool :ensure t :config ...`. The `:ensure t` clause will cause it to fall back on `package-install` in the event it's not bolted into your local Cargo-style-own-the-world-and-dont-break repo mirror. Disks are big, it's like the right default now.
But if you've got a package that's mis-specified in terms of where on `github` or wherever that `striaght` is supposed to find it, you can find it tapping `elpa.org` on the shoulder at a time when the UI thread is also doing blocking network IO.
The `Messages` and/or `straight` buffers will have warnings about this, so it's fixable to find the offending package and it'll stay fixed, but in fairness what you're describing is possible.
> There’s a market for people who want something snappier than VSCode but less labor-intensive to set up than emacs, and I wish them luck
Basically AstroNvim or Doom Emacs. Not a huge fan of Doom Emacs but AstroNvim got me to drop my main editors (Sublime + Atom) and I basically only use AstroNvim.
One enormous advantage of Nvim is that I can run it anywhere. I run it on a Linux machine, a Mac, and a Tablet (w/ Termux) extremely easy (just clone my dotfiles, install nvim, and that's it).
>>But emacs 29 with the right flags and a tuned GC (no one does this! it’s got a heap-size from the 80s!) is just as snappy and has more amazing packages than VSCode.
That explains my whole experience with Emacs. More time is spent in making Emacs awesome than actually doing my work.
Most language modes have a `foo-ts-mode` now that uses `tree-sitter` to get a perfect* AST in the absence of errors, and almost always the most reasonable one in the presence of errors. `tree-sitter` is amazing, a real feat.
It ships with 29 and it's work to find `brew` or `apt` or `nix` or anything giving you less than 29.1 these days.
Have you tried a recent version? 29.1 is dramatically more accurate on everything from syntax highlighting to indention than VSCode (IMHO, they're probably either using or working on using `tree-sitter` too, those are serious people) or JetBrains stuff (they're working on their vscode clone more than IntelliJ these days).
I only use `nvim` for commit messages and stuff the last few years, but AFAIK it has all the `tree-sitter` stuff too.
If it's the sloppy naming of `c-basic-offset-this` vs. `py-indent-that` or whatever (I don't even remember), try a good baseline distro like Doom and tweak from there.
VSCode has always felt incredibly slow to me, even compared to e.g. pycharm. , which I have always assumed of be otherwise roughly comparable. VSC’s lag in basic code inspection and linting became so annoying I had to switch off it. We’re not talking seconds, but maybe tenths of second lag, for everything at all times. I understand plenty of people love VSC, but honestly I have never been able to share that enthusiasm.
Yep. For some reason, suddenly, vscode became painfully slow on my decently spec'd machine; 3 to 7 seconds per keystroke just to analyse the file to show intellisense tooltips.
Yes, that was in seconds per keystroke.
The irony is that I moved from neovim to vscode because setting up intellisense in (neo)vim was always a hassle and never worked quite well. Pylance seemed too attractive not to give it a spin.
Now the lag has as mysteriously diminished, but still vscode is very far from being as snappy as (n)vim.
Ive been wanting a syntax-tree-viewer for months, to help me learn functional languages where figuring out what is even going on syntax-wise in the exmaples provided by tutorials keeps being an issue for me.
Does anyone know of a way to see a syntax tree for any given snippet of code for any given language? I'd try Zed, but I'll have to wait for Linux support.
I am not sure what debug syntax tree mode does in Zed, but if it's about tree-sitter generated syntax tree, you can see that in Neovim or Emacs (assuming you have major-mode/grammar loaded):
> older editors that are doing syntax highlighting via regex.
I mean, Emacs, which can probably be considered the oldest code editor at this point, got built-in tree-sitter (which is what Zed uses under the hood) support in the last release. So it's not really related to editors being new or old
It's not just tree-sitter that makes zed feel snappy.
If you're using a reasonably fast language-server, which rust-analyzer apparently is (I didn't know this using vscode), the autocomplete & intentions feel instantaneous.
I think the team has learned a lot from previous editor implementations (they were the core team of atom that was notoriously slow), and so they've had an opportunity to do a lot of stuff right.
FWIW they also are the team that originally wrote tree-sitter.
The quickness feels more like it's in the core of the editor. I was shocked how much it impacted the editing experience when I tried it in early beta.
> If you're using a reasonably fast language-server, which rust-analyzer apparently is (I didn't know this using vscode)
Some more anecdata to back this up: initial workspace load in VScode I can watch RA tick through its progress. Clean and boot up Zed and the same process is so fast that it’s almost unbelievable.
I've been impressed with the C++ clangd language server snappiness in Zed compared to both CLion (my old favorite) and Emacs (apple of my eye for 2.5 yrs).
I always thought the major slowness was coming from clangd itself, so I'm surprised and impressed to see that Zed appears to be quicker on this front. I might be using Zed as a 'second opinion' editor because of this.
However, now I'm used to the infinite customizability and coziness of Emacs, it's going to be hard for me to move across to Zed permanently.
As I understand, tree-sitter implementation in Emacs is currently more like a foundation for development / adoption by plugins, it's not really usable as-is today.
Your understanding is outdated/incorrect. Already a number of major modes have corresponding ts modes and I use them daily. This includes C++, Rust, Python, yaml, toml, json, Java, typescript, dockerfile and cmake.
True not every major mode has treesitter in the official release but I’d hardly call the above “not usable”
Emacs already had tree-sitter before 29, just not built in. You can do syntax highlighting[1] and structured editing with it.[2] What is missing for it to be usable in your view?
>>I have reasonably fast machines, so we're not talking about the difference between 5ms typing lag and 500ms, but it's still pretty surprising.
I really envy people who can sense passage of time between intervals of 5 milliseconds and 500 milliseconds.
My sensibility begins over a second. And to be honest even that is least of my issues. Same with start up time, I restart the IDE only once in a few days. I thinking spending a second or two extra for it is any where in the ball park of what I would call wasting time.
I use this extension: https://continue.dev/ their docs are pretty good, but it's also evolving pretty rapidly. For example, you no longer need to run the continue server yourself, it's entirely self contained in the vs code extension. I believe the docs still refer to how to run it manually.
I work for a pretty conservative company re: GAI, and the ollama + continue combo made it through legal.
I switched _to_ VSCode because Sublime was occasionally so slow at times that it was unusable. It was very fast 95% of the time, but then I'd `git pull` on a very big repo and my machine would become unresponsive while Sublime did...something.
Sublime Text 4166 has some major improvements in that area, previously ST would check every file in the side bar for changes in order to update its index but this is now done incrementally.
This is very strange to me, because I have always used Sublime Text for super-giant corporate repos, and always found it snappy compared to vscode. Wonder if it was a specific usecase or something...
Every little while, I try to make vscode my main editor, because I enjoy all the features. I always switch back to Sublime just because VSCode's slowness bugs me.
Haven't really tried Zed in earnest yet, though, because of no custom LSP support.
I think on Linux at least it freezes due to watching too many files using inotify to update its sidebar. I wish there was an option to disable that. Same thing happens with Sublime Merge. It’s unusable on giant repos.
Not in 15ish years and at the time I think I had a 5400rpm hd, so that probably limited any perception of sublime being noticeably faster than other editors I was using (geany, kate). I don't doubt that sublime is faster than vscode today, but the vscode ecosystem is a pretty nice place to be and probably worth trading some speed for, especially on a nice mac where the trade off is probably small. Zed might be even better however.
Or vim, I suppose. It's remarkable how slow VSCode actually is, but I still use it because I hate configuring vim and packages always break when upgrading, it's honestly worse than npm.
Updating vim packages can be a nightmare if you don't vet your plugins and generally install stuff "just in case I want to try it out later". If you aren't using a plugin with any regularity, delete it.
I don't know what vscode's plugin architecture is like but I imagine it's just that Vim's is extremely powerful and, you know, with great power comes great responsibility and all that. Unfortunately, many vim plugin authors are not responsible, or don't realize they aren't being so.
I’m still using Sublime for the references. Without fail I can load up any version of an application including older rails apps sub 3 and it instantly has context. I can easily jump to definitions and even on hover will show me where they are being used. In the context of foos.bars.baz I could easily infer where bars is defined. Seems to work across languages, doesn’t matter if it’s js, ruby, in an erb file it just works. Not to mention the flavor of vim feels great.
VSCode complains and I’ve got to hardcode an alternative absolute path gemfile for Shopify’s LSP to work. It also feels clunkier, even on powerful machines.
If Zed can give parity with Sublime on the references (I just tried and it did not seem to find any references when clearly bars was there) might be an interesting change. Considering it’s open source now I’d happily switch as it does seems super quick.
I might also be old-man yells at cloud and the copilot integration doesn’t appear important yet, but current sublime support is pretty poor for it and seems like that’ll be a thing more and more.
That’s where I landed after v14 with LSP support came out. It’s so nimble! And extensions with shell scripts and CLI programs is right in my wheelhouse.
> GPL for the editor, AGPL for server-side components). GPUI, the UI framework that powers Zed, will be distributed under the Apache 2 license, so that you can use it to build high-performance desktop applications and distribute them under any license you choose
Interesting choice on licenses.
—-
I’m been super happy with Zed, my main requests (and I’ve sent in this feedback to them or contributed to existing GitHub Issues)
a. Window Size & Position doesn’t persist after closing Zed.
b. I constantly run into Language Server errors
c. Alabaster use to work as a theme, doesn’t anymore. Would be great if you could import VSCode themes into Zed
All above has tickets open.
----
Hope these small things get addressed because I truly love the elegant UI design of Zed
For those who haven't used Zed, it's the first GUI editor I've used in 25-years of development that wasn't distracting.
It's hard to describe how much more focused I am when not distracted with a Christmas tree scene of icons, menus, colors, etc. like you see in other editors.
Zed is very calming, due to its focus on not having distractions. Give it a try if you haven't.
The licensing choice is smart. "Permissive" licenses permit closing off something built largely on the work of the open source community.
Zed is made by the Atom guy and the Tree Sitter guy. Atom was MIT licensed. I wouldn't be surprised if he's thinking the MIT license is the path to VS-Zed.
This tripped me up as well, you need to set up the language servers in your settings json. There are examples in the Configuring Zed section of the Docs area on the site.
Funny you mention that because I felt Zed had _too many_ buttons & icons for things compared to Sublime Text. I feel they could do without the dedicated stuff for GPT/Copilot and others, and sweep them away to the command palette.
I opened a random python project on my machine in Zed and it automatically loaded up an LSP for python. It looks like it's using the same one as my emacs uses (pyright), but it presents completion choices in a not particularly useful order. Typing `os.p` gives me for instance as completion choices:
Is pyright just giving Zed all the possibilities and it's up to Zed to rank them? I don't know the details of editor/LSP integration. lsp-whatever in emacs ranks these choices in a reasonable order.
I’m also curious about the answer to this! I noticed similar behavior when opening a Typescript project. Enjoy the low latency, but I’d also appreciate accurate/helpful autocomplete suggestions.
Yeah, I noticed while writing Typescript that it was suggesting `String` above `string`. In Javascript and Typescript, there are almost no situations where `String` is the correct thing to use, so this choice surprised me.
I tried it out today, and in all fairness it is a very attractive and pleasant-to-use editor, but it felt like it was missing a lot of the little details. Things like suggestion order, or being able to open terminals side-by-side, or showing quickly where all the type errors are in my project or even a single file.
The collaboration tools sounds really neat, but unless I can convince colleagues to use it with me, I don't have anyone to collaborate with! For that, the little details are hugely important.
That said, it looks and feels wonderful, it's a very elegant tool.
Hey! Thank you for this cool editor! However, has it been written with cross-platform support in mind? Otherwise, porting from mac to linux could be rather painful and time-consuming... Will there be an ETA for us linux users? I saw it on the roadmap [0] but without an ETA.
"We will let you know when windows support is under development. The Zed team isn't taking feedback on when we start on windows support, nor why we didn't start with windows.
The amount of people on each platform isn't relevant, as it starts with the base assumption that more people = better to the Zed team. Growth driven development unlikely to lead to a resilient, high quality product long term."
Haha, well they will get lots of feedback they don't want to take.
A company like Jetbrains can do this to an extent due to existing products fueling long term R&D efforts before they bare fruit. Interested to see how this outlook holds up as the runway burns and investor pressure increases from that cool 12.5m they raised last year.
From what I can find, the used OS of software developers is 61% windows, 46% Mac so this argument has no weight on the considerations the team should make Å— it’s essentially a toss up.
Even if it weren’t a toss up and significantly more devs used windows, deciding to target windows based solely on OS market share is one of the worst things a small dev team with finite resources could do. It’s just not a compelling argument at all.
I think linux + osx combined is probably developer majority. I'm assuming most Windows development is .NET or Unity based. Over the past nine years, I've worked primarily with Node, Ruby, and a smidgen of Java and no employer has issued Windows machine. It's in sharp contrast with the start of my career where I was in VB6 and C# where I only worked with Windows environments. Could be confirmation bias, though.
Windows is a great platform for developing any language or platform, not just .Net. Personally I think it blows macOS and its dated tools out of the water in any category
Stackoverflow found that Linux and macOS is the slight majority for professional use and the minority for personal use with developers (with the caveat that the total professional use of all categories is >100%)
Just curious what tools are you missing on MacOS or Linux that are found on Windows?
I have been on MacOS the last 8 years or so but before that was all Windows and Linux. I prefer MacOS now but curious what I am missing. My colleagues that are using Windows machines all use WSL with dev containers so really just using Linux under the hood.
My experience is that pretty much everything is cross platform these days. I don’t do .Net or game dev though. Just Go, Java and some NodeJS these days.
Yeah, but with Windows GUI along with fractional scaling, device, and etc support haha. I've developed on both bare Linux and for years in Linux VMs. Saying Windows+WSL is "really just linux under the hood" does this setup a massive disservice. It makes me smile every time I type WSL into the terminal lol.
Yeah Linux desktop environments are a bit of a mess but OP specifically called out MacOS as well. Once you add Rectangle it is a solid DE in my opinion. Also it is Unix/linux compatible so kind of the same beast with good GUI/hardware and *nix terminal.
Was just curious what Windows only tools I am missing out on.
I agree. Long time Linux engineer who prefers Windows to OSX as a desktop OS. Currently use a Windows workstation with WSL and VSCode as a daily driver. I feel this currently gets me the best of both worlds(Windows/Linux).
It is, but barely (windows is @ 61%, macOS is at 46% based on the numbers I found). And even if we ignore that, number of potential users is far from the only factor you’d consider when choosing the first platform to target for a new project. And then, ignoring that as well, just because an OS has more potential users in total, that doesn’t mean it has more potential users of your editor.
There’s lots of reasons to not choose windows as a first release platform, and it being ‘the most common development platform’ isn’t a compelling one, especially for a small team with finite resources.
These days that's a lot less of a concern than it used to be. It is a lot of up-front work to facade e.g. Metal, Vulkan (and even D3D12) but it's much much much easier than back in OpenGL vs D3D9/10 days.
Most of the general concepts are more or less the same across them all these days. The "shapes" of the APIs are very similar.
A texture is a texture is a texture. Same for a vertex/index/uniform buffer, vertex/fragment/compute shaders (notably not geometry, but you can just use compute), etc.
I'm not in the graphics world at all, but I read this lively discussion last week (post [0], discussion [1]) that made me think the situation was pretty messy. Is it maybe not as bad as that makes it appear?
I'd recommend wgpu. It should be a fairly easy migration from Metal, as Apple has been actively involved in the design, and there are similar concerns for developer experience (unlike, say, Vulkan).
Unfortunately, graphics is in many ways the easy part. Really excellent integration with system menus, preferences, keyboard, input method editing, all have more variation across platforms and a still-evolving story of solid Rust abstractions. Lately, we've decided to try joining forces with the winit project to see if we can get those problems solved well.
“Please don’t do it like a 4x4 truck, do it proper, like a bicycle” — there is some overlap between those things, but for people who are actually making full use of the former, the latter is not a useful suggestion
I'd suggest that the people who make "full use" of vsc-over-ssh are satisfied with vscode, so it would be unwise to target the full featureset.
More generally, targeting another project's complete featureset is often a great way to get mired down in the wrong details. Unless you can afford to do a proper cleanroom -- then, you'll be able to at least match the performance and useful abstractions used in the original.
> I'd suggest that the people who make "full use" of vsc-over-ssh are satisfied with vscode, so it would be unwise to target the full featureset.
Remote SSH + Dev Containers and their seamless integration (even stacking one on the other) are the only features that keep me using VS Code. I would love to see the full implementation of these in an editor as fast and light weight as Zed.
How exactly are these features present in your workflow? I honestly struggle to think of when I'd ever use this.
Fwiw, I do a lot of infrastructure-as-code, full stack, and systems programming.
I usually have a split screen (editor | terminal) or two terminals on the side, and exec into a container, or use devenv.sh.
If I _really_ need to modify files in the container as I dev and a "make" doesn't cut it, I usually just run podman with a -v mount. Similarly for remote machines w/ sshfs, though I try not to.
Devcontainers are amazing for getting a consistent environment set up on multiple computers on multiple operating systems. I was in this situation recently with a new colleague who was using a very locked-down Windows computer, and it was really convenient for the "you must install" list to be Docker and VSCode only. It's definitely not ideal - it adds overhead on Windows and Mac, there's occasional networking issues, and I don't have all the creature comforts of my usual shell - but it's very convenient for what it is.
Similarly, editing code in-place over SSH rather than rsyncing back and forwards is very useful for certain types of embedded development. I worked for a while on sensor systems where a lot of the logic was handled by a Python webserver that couldn't really run without access to the sensor data it was using. Developing entirely locally was therefore difficult, but developing on the machine was also painful because it didn't have the right tools. So we'd work locally, and then copy the Python files over every so often and restart the server. At the time, I don't think VSCode's remote stuff was working as well, but I believe now it's a lot better and could have handled that situation well - edit everything in-place, run it immediately, but still have the power of you local development machine available to you.
> I'd suggest that the people who make "full use" of vsc-over-ssh are satisfied with vscode, so it would be unwise to target the full featureset.
Wha...? This is a killer feature. I'll put up with a lot of crap to use it. That doesn't mean I wouldn't switch to something nicer in other ways if it also offered this killer feature.
It's hard for me to understand why there are IDEs under active development not trying to offer this feature. It is so much better an experience to have the network split into the proper place: between the UI and the heavy computation. Having the UI too far away undoes whatever responsiveness work has been done and more. Having the heavy computation too near means it's hard to develop and test the far environment, take advantage of its compute, etc.
> the people who make "full use" of vsc-over-ssh are satisfied with vscode,
almost all my coding these days is over vsc-ssh. if zed supported sshing into a remote host and into a docker container as seamlessly as I can in vscode, Id switch immediately. The performance of the ui in zed is so much better. I'm a bit sad I can't switch with its current feature set.
The former is bad design, it has no gradual fallback and works in a terrifying way. It pulls binaries from MS servers and runs a headless vscode instance on the remote machine.
You are dead in the water if you target a machine/architecture it doesnt support.
You want to target FreeBSD? Linux on POWER or RISC-V? An old (ARMV6-TDMI) Raspberry Pi? Sorry, no remote work for you.
I'd very much prefer something that works seamlessly in 95% of the use cases and makes me do some work or look for an alternative for the remaining 5%, than something that makes me do work in 100% of the cases.
Bikes likewise have advantages that 4x4s lack. If you mean to imply everyone ought to be satisfied with the 4x4, and not ask for a bicycle: no. We can have both.
I don't think I would like that. I haven't been impressed with the experience of running GUI apps within WSL. Weird window controls/resizing, compositing, and font rendering issues would be a hard blocker from using this as probably the second most important app on my computer.
Their UI framework is also rendering via the GPU, so it's more likely they'd wrap it in a GTK/Qt/whatever window and then just render accordingly. You don't need GNUStep for it since there's little "mac"-isms you need to cover.
Speaking of Linux, are there any other editors we can use today? I tried to look for one recently but couldn't find anything. (edit: I mean collaborative editors)
Precisely, I'm looking for collaborative editors. Thanks for the suggestion, I hadn't seen this one.
Unfortunately, it seems that this plugin only shares the editor window and does not keep a local copy of the files. I was wishing for something that would let both sides run & compile the code during the editing session, without having to stop for git push.
https://github.com/MicrosoftDocs/live-share/issues/3524
Jetbrains allows both users to execute / debug but it’s all on the host machine.
What you’re talking about sounds like you should do an rsync script that watches file changes, but seems like a good way to constantly be breaking the other persons workflow.
git solves these problems- feels like the right tool. But maybe I’m missing something about your workflow
Looking for collaborative editors in particular. Two people editing at once, on different computers. Each person has their own copy of the file, on disk.
Authoring Latex documents. I want a non-browser alternative to Overleaf. When the paper is close to done, me and my colleague are on voice call, editing the same file. But we both want to be able to compile it to PDF, to be able to see the figures.
Gobby is mostly dead but still works. Win/Linux; macOS builds have gotten harder now that it's been about 3 years since the last release: https://gobby.github.io/
EDIT: It's apparently on Macports as of quite recently, though the port health for recent macOS releases looks bad.
I was particularly looking for collaborative editing; two people editing the same file at once and each has a local copy of the file so they can run the code. Perhaps one could use a normal editor for this if there were some underlying command-line tool and/or plugin to do the actual file synchronization. Do you know of any?
AI helpers, Chats and Calls??? Microsoft log-in, Commercial plans, Telemetry? Seems like an overload of everything for an editor - more distractions not a new way to code. Def sticking with good old nvim for some longer time, no better editing experience engineered yet
It is not really needed, but this one is definitely here for the sake of announced commercial plans. Maybe it is faster than vscode, but it is still not fast enough like most of based terminal editors. I just made my benchmarks with 500MB of dev/urandom bytes to open on clean nvim and zed. Nvim opens it in 20 seconds while it takes 30 for zed, so I can say there is at least something inside the core engine which works 50% slower that of my champion's. Non of the landing page benchmarks at zed features a fair competition with really fast and great editors.
Editor's main purpose is to edit, and i don't think that stuffing anything else in the basic box without user asking for it is necessary. You can always pick a plugin if you need AI or collaboration or suddenly hundred species of messenger is not enough and it MUST be inside you editor instead of a usual swipe to an adjacent screen.
The performance we're talking about here is not startup time, but responsiveness to keystrokes. Fast startup should also be a goal, but this is primarily an editor for doing development work with, and for me at least those kind of editors stay open all day (or more) for each project I'm working on. I use a very lean vim configuration for other use cases.
Really nice and fast (which is the reason I’m still using Sublime instead of VS Code). I’ve only found one piece of functionality that’s dramatically slow compared to Sublime so far: selecting all of the current highlight. In my current file I have 2,396 occurrences of `<span>`. Selecting all of them in Sublime (with ctrl + cmd + g) is certainly less than 200ms. Doing the same in Zed (with cmd + shift + l) beachballs my machine for around 5 secs.
For most code work that’s probably not a situation to optimise for, but I often work on large markup documents.
Zed developers, if you read this, please get inspired by Cursors "Fix it" button that you can click on any error. It simply starts a new chat with the code context and error message, suggesting possible fixes.
I'm currently learning Rust and this is such a powerup that I honestly wouldn't know how to learn anything without.
> Zed supports GitHub Copilot out of the box, and you can use GPT-4 generate or refactor code by pressing ctrl-enter and typing a natural language prompt.
https://zed.dev/
Seems like that ship has sailed. Maybe it's a plugin already or could be in the future, but that's not on GP's suggestion.
This made me chuckle. Then at the same time I find it disheartening that privacy is currently seen as an enthusiast position rather than the default. I really dont mind the idea of these anti-privacy things being included as long as I have an option to turn them off before running Zed for the first time.
To use them you’d need to already have a copilot subscription, and the GPT-4 thing apparently requires you to deliberately invoke a command. I’d say it’s pretty safe by default.
I’d say the ship is in port but ready to sail. Before this does anything you have to provide your own API key. So it’s off by default. It’s just the button that is there.
Beyond "is it on by default" (due to data privacy concerns) whether something belongs in a "core" editor or in a plugin is a whole separate can of worms - moreso if the core editor starts shipping plugins. There is someone who reads that agreeing it means the editor should be just a plugin store and if you want font hinting, syntax coloring, tabs, terminals, smooth scrolling, and so on then download them yourself. At the same time there is someone who reads that agreeing it means the editor should have anything possible related to working with text but if you want to play a video while you code the plugin system should allow you to do that in a pane. Neither are really right or wrong about what should be a plugin, it's really a matter of what the tool wants to optimize for out of the box.
I've been following Zed for quite some time now and happy to see them follow through with the OSS move.
I personally don't like for my editor to send out any kind of external requests at all, and this is actually what keeps me on vim as my main editor. I also don't like limited login options
It would be cool to have a version that's just a stripped down Zed, and if need be you can install the extra stuff as plugins.
Zed was the name of the Zortech C++ IDE back in the day!
It disappeared, though, because the author (Phil Murray) disappeared and we didn't have a license to continue with it when Zortech was bought by Symantec.
It's too bad, it was a nice editor. I never was able to find out what happened to Phil. He was an excellent programmer, and an all around pleasure to work with.
I had the pleasure of working with Nathan Sobo when he joined Pivotal Labs ~2008. He had a burning desire to create a transformative text edior/ide back then. He is bright, not only in an intellectual & talent sense, but also in an energetic sense. He is also compassionate & genuine person. I'm very happy & inspired every time I see his endeavors progress.
Knowing Nathan has been one of the major catalysts for me to improve the art of software craftsmanship. Sometimes you meet amazing people in life & I count Nathan as Amazing in many ways.
I credit Nathan Wilmes for showing me how to speed up my development techniques & decision making...Mainly by observing him work during our pairing sessions. It's interesting how one can learn effective practices that are difficult codify with conscious thought & written or spoken language...Yet the human observational systems & nervous system are effective in working with these complexities. It's almost like how a GPU frees the CPU from processing graphics or linear equations.
I usually try to get into an IDE from vim every 6 months/year and Zed is the first IDE when I actually came close to being successful. Usually what kills me with IDEs is distractions, things getting in your way, and responsiveness. Zed is the "cleanest" IDE I've found so far on many of these points. Not totally 100% (pyright being the most prominent annoyance I'm trying to turn off right now) but not enough to make it give up in just a few hours!
I only joined Zed, the company, a week ago, so take this with a grain of salt, but:
I've never really used real-time sharing (tried VS Code's Live mode and other apps) for longer than just 1-2 attempts, but at Zed things are different somehow. Everybody is constantly available in a channel and people just hop in or out. In the last 7 days I spent 3-5hrs every day pairing with others, using Zed's live mode. No video, only audio and sharing code in Zed (no video was weird at first, but now I think I'm starting to get used to it?)
IMHO it's a combination of culture and technology, but when the mix of that is right it really feels game-changing.
I think what’s nice is that it’s frictionless. If I can just click a button to see someone’s editor, then there’s no dance with sharing screen, I can open files I need to see, etc.
With standard sharing you always end up trying to direct people to do things instead of just showing them.
That may depend on what you're trying to do. If you are figuring out something tricky, then lots of quiet head down time is what you need. Every interruption hurts when you are concentrating.
However, a lot of the time is just figuring out how to glue together multiple systems. Being able to pull in various people to interface little bits is priceless. There is no flow here, only collaboration.
As a senior engineer, it really helps when helping junior engineers. When I'm on a call and we're sharing screens it gets really annoying having to tell them "no, go up. Go up, all the way up, the button at the top. THE TOP"
This. The worst thing is that I get mad at them for not doing what I want them to, faster, when the tooling and my crooked explanations are at fault, not them. Not that I mention that out loud, but sometimes I feel ashamed for those feelings.
Having real time collaboration that works so seamlessly as Zed apparently does would really get me to consider switching IDEs!
I'm using VNC for coworking/mentoring sessions and it works very well for us (it's lower latency, higher frame rate, sharper picture than typical screen sharing, and allows both/all sides to interact with the session and use the mouse pointer to point to things). We're using Linux on both the server and client side. There's a full desktop on the remote side, to fully interact with that (to have key bindings go to the server instead of the local desktop) the client needs to be made full screen. Latency isn't really an issue even between central US and Europe.
I've automated the set up of a Debian server for the purpose using[1]. There are some details on how to run it and how to set up the client side here[2] (probably slightly outdated). Feel free to mail me about it.
You can share terminals in live share and even share servers etc. I’ve had no issues and I’ve built multiple teams that are very collaborative at this point. Dozens of engineers
This looks really cool, but I have a suspicion my employer would be against external connections. Any distant plans in the future for an enterprise/onprem license?
When a person tells another person "you need to do X" or "I did Y and it didn't work on my machine", they sometimes ignore important details due to their own curse of knowledge. That's when I need a shared terminal/editor so that both parties know exactly what is happening.
I don't frequently share terminal/editor but occasionally need to do so, maybe a handful times a year. Most of times when I tell a new colleague "hey, connect to my tmux socket by running this command / opening this link", they were amazed and thought it was magical. I am not a Zed user but it's always welcome to see people making effort to make pair programming easier for everyone.
Now, I don't know how I would feel if one day my colleague send me a sharing session link that can only be opened by a particular editor.
The benefit as a viewer is you can drive navigation on your own without asking the presenter to go to a certain spot. So the presenter can be discussing a problem or a function, and you can investigate/gather context without interrupting them, among other things.
I personally can't live without it, and am almost annoyed when folks aren't using VSCode/have no way of doing a live share. Screen share sucks.
It sounds nice in theory but I've noticed I cannot focus when somebody else is looking over my shoulder. It could be useful as a way to occasionally show somebody around the code base though.
It is such a game changer for me. Pair programming or multi pair programming if you will, suddenly has 0 barriers. Including others in your session with no overhead changes things completely. Before I would rarely include others in my processes because it cost too much. Now I'll do it whenever... Switched from sublime and got my coworkers macs just to be able to use zed and work in this way.
It is extremely important it works fast and fluid, and zed is the only one I've tried that nails it. There are still a few things that needs tweaking wrt. undo history, but I'm sure they'll get that to feel intuitive in the end.
I think it's mainly the intensity of the sessions. You've got someone on the other end of a call so you don't really chill and take it easy. Creates some urgency to get the thing done and you're just a bit fried by the end.
I've greatly enjoyed pair (and trio) programming with friends using it (vscode's version of this) on side projects. For the most part I'm not sold that pair programming is an efficient way to program, but it's a nice social way to program.
If you're going to do it, especially remotely, it's really nice to have your editor support it instead of just using screen sharing.
I've never really pair programmed - I've worked on joint projects, of course, but collaboration occurs over discord/github. I'm intrigued - what does this actually consist of? Does one person write one function and another write the next while you chat about what needs to be done?
We tend to work on one 'thing' at a time. Sometimes that means we're both/all on the same few lines of code writing a new function - at which point one person is usually doing all or nearly all the typing. If you then need to go define a constant somewhere else in the code or something the other person does that so you don't have to pause. Other times one thing involves code in various places because you're doing something like changing what arguments a function takes, or fixing compiler warnings, and we do spread out and handle instances of the thing we're fixing as we come across them.
Lines of code per second it's closer to one person than the 2 or 3 people involved, but the quality of those lines definitely improves a bit as people spot each-others mistakes/less than optimal choices. (Edit: But I'd emphasize we're doing this for fun, not to maximize productivity)
ime, real time sharing works best when there’s shared context. otherwise it’s essentially an artist rendering a public performance in the presence of an audience. unless you derive some joy from watching people code it could get frustrating real fast.
Honestly, it's pretty cool in my experience; in VSCode everyone gets their own "client-side navigation" so while you are typing something, your coworker can go look up a function for you in another file, etc. Even though they exist in "your" instance, they still get Go To Definition, Find References, type-on-hover, etc.
Realistically though it doesn't replace every instance of screen sharing, for me. It's also cultural to some extent I guess. For things like debugging or being intro'd to a new codebase, I think it's great, though.
For example, in vscode if the sharer start a http-service the port can be port forwarded through the session so anyone in the session can interact with the service.
Maybe I'm opening the lid to the industry's best-kept secret, but every team I've worked on used SSH with a shared tmux session for pairing. There are probably more complex tools that cover every use case, but I never encountered much demand for one.
I just tried Zed and it is definitely really pleasant to use. I'd need at least a Java LSP to use it for real (and preferably general LSP support so I can code in my own programming languages with it) but I can see a lot of promise. If they ship a Linux release so I could use it on my desktop too they could win me over from Sublime Text/IntelliJ.
About 70% of the programming I'm doing right now is in my own language. But like, my full time work is programming language research so I accept that's an uncommon use-case.
The polite thing to do if you have a feature you think would make a FOSS project better is to submit a feature request first. If the maintainers decide they won’t support your feature quickly enough or at all, then fork it. GP did the right thing here
Their language support story is extremely weak right now (no support for Java, XML, SCSS, etc.). An editor should focus more on editing capabilities (it is editor after all).
Also, no option to disable all the collaboration and AI features leaves a sour taste for me. Hopefully, they'd improve on this later.
This looks really nice, but I am confused why the install is over 350mb, including a 300mb binary. I thought Electron-based stuff was big but this is even somehow more!?
The core app itself is 100% Rust, but it supports integration via Microsoft's Language Server Protocol[0][1]. In practice, this means you will be running daemons locally for each language, which may be written in any language. Often these daemons are written in NodeJS since the reference implementation[2] is in NodeJS.
It doesn’t follow that a language server written with JavaScript and run via node will bloat the binary by hundreds of MB. Are they bundling a node runtime too?
Maybe if they are embedding dozens of language servers and runtimes it could bloat the binary, but I assumed the extensions and language servers would be downloaded on demand.
But a rust binary by itself shouldn’t be that large. LSP is just a simple json protocol, so parsing it doesn’t require hundreds of MB.
When I opened a Rust project I think the status bar said it was downloading Rust support, which supports your hypothesis. (Also it doesn't make much sense to bundle these things when they are all making new releases at different schedules.)
Hey, Zed developer here. Indeed we do not bundle the LSP binaries into the final binary for the reasons you've stated; and I do agree that the binary is kind of big, though at present .dmg compression gets us a long way (as the .dmg itself is ~115Mb).
Right now we ship an universal binary, so half of that size is essentially unused:
Interesting to see how they are gonna approach integrating plugins/extensions system, because this is likely gonna be one of the major factors affecting adoption and ecosystem growth.
Helix devs, for instance, lean towards a Scheme-like implementation. [1]
Tried Zed some time back, the performance was pretty good as compared to VS Code, though gave up when they chose to implement other fancy / collaborative features before implementing the most basic requirement from an 'editor' - Search and Replace. Contacted them but there was no response. Looks like they have replace now, so worth trying again (still hard to give up all the features of Sublime Text yet).
It's interesting that the #2 selling point on the landing page (the Copilot integration) relies purely on Microsoft's goodwill. There is no official API for Copilot but instead a (non-standard) LSP implementation embedded into the proprietary Copilot Neovim plugin. Zed seems to trust under the hood that a release of the plug-in exists on Github, pulls the minified-js language server from there and integrates with that. The minute MS decides to pull the plug on the Neovim plugin Zed loses the functionality as well.
I wish we had a proper API to interact with copilot but it seem that pulling everyone else except the dedicated VIM and JetBrains users into VSCode land seems to be more in their interest.
It is not paid product for non-commercial and open-source products.
The get massive amount of data if you use it in VSCode, and because of this, it is pain to ass to use on VSCodium for example. Your data is the currency, and other editors are good on limitin that.
If they wanted non-sanctioned editors integrating Copilot they wouldn't have taken the extra steps of implementing an obfuscated LSP but provided an API or a plain client library. At best this could be a pet project by a Neovim fan GH employee. Either way I wouldn't expect long term (if any) support from this thing.
> the Copilot integration) relies purely on Microsoft's goodwill
Given that the founders and several employees are former GitHub employees, I have to imagine they know how to do this integration in a proper & officially allowed way.
To any of the project members here: viewing the zed.dev home page using Safari on an iPhone shows text being cut off. I can’t seem to scroll to the right either to view the remaining content.
I don't understand this comment because American/European governments (where I would guess most HN commenters are from) are almost certainly doing the same thing according to what we've learned about for example the NSA. Maybe you trust your own government more.
Most westerners I've spoken to think that even when a Western government does bad, there's probably good intentions behind it or at least more noble than any non-Western government.
On the flip side, even if you believe that, there's an argument to be made that if you live in the West then the Russians/Chinese care way less about you than the Five Eyes.
Few decades of not starting a war is a very weak qualifier. Let's start with "100 years of not participating (!) in wars, not even in a supporting role, while respecting people's human rights, democracy, freedom of speech and privacy" and then the country has some kind of backbone to lean on when offering it's enlightened perspective.
There aren't many countries that would qualify (Sweden fails because of the Snowden case) and I suspect your country is not one of those. What country are you from?
It's all about risk assessment, risk analysis, scenarios, probabilities, threat vectors, worst/probable/best cases of cost/benefit, etc.
What are the chances that Putin wants to get to you? How probable is it? Does he have better options than through seeing your browsing history? What happens if he actually gets it - what's the worst that can happen?
It's really all a big nonsense about nothing. You are most probably just an ordinary human, just like me. We are not even worthy of such an attention. Be realistic and know your position, role and pecking order in the world order. Be a cog and live peacefully. :-)
Yeah, I just think that after Snowden (but before him as well) it's naive to think that we in the West are free from being spied on. The methods are more refined, I'll say that.
As a Finn, both American government and Putin have the same amount of energy put into trying to kill be. As in zero energy. Because if it weren't zero, I would already be dead. And it would happen kinetically, not through capturing my browser history.
And anyway, it's a bad comparison, there is a conspiracy that, say Google or Firefox is collaborating, but it's very clear that yandex is an arm of the Russian government
I might be missing something, but why is it clear to you that Yandex is an arm of the Russian government? Yandex has been critical of the Russian government's war on Ukraine [0] and faced consequences because of it which signals to me that this isn't the case. Yandex made plans to exit Russia because of the company's disagreement with the government even. [1]
As someone living in the west, I'm disappointed by our governments' hypocrisy over the massacre in Palestine while supporting Ukraine's right to resisting an invading force. Our governments' support of the massacre in Palestine (even dealing out death to peaceful resistance, as we can see by the bombing of Yemen when its people merely made economic trade inconvenient to protest the massacre) made me think the people in charge of western countries are no better and probably even worse than dictatorships such as China and Russia. (Western governments are just better to their own citizens apparently.)
It can be argued, that Yandex didn't do enough to fight the Putin's regime, but:
1. How realistic is to expect something like this from Corporations? How much have American corporations fought against American terrorism or attacks on freedoms (e.g. of speech)? Exactly.
2. It absolutely does not mean that they were an arm of the government. Among the tech giants, even without the Russia caveats, Yandex was definitely a good guy, with good ethics and core, with people who cared (and who helped a lot in personal capacities).
A very telling sign is a list of Western countries that participate in the Yemen war, but want to keep it a secret from their citizens. Yes, they fully know that what they are doing is wrong. No, they don't want to stop doing the evil, they only want for people not to know.
This is a pattern. Remember illegal CIA prisons all around the world, torturing people?
Democracy often means that bad deeps must be better camouflaged/hidden, not that thr actions and people doing the deciding are inherently better.
1. I live in the EU and I would worry much more about my country/Union spying on me than foreigners.
2. In the same token I trust Yandex much more than Google. And yes, I am competent enough to make that assessment. The same goes with using the world's best antivirus in Kaspersky vs. using something weaker from Five-Eyeish+block.
3. Yandex Browser is technically good. But the main killer feature for me is having the address bar on the bottom of my screen. This should just be a common-sense default for mobile browsers, yet it is't.
Worth noting that, like many immediate mode UIs, it seems to not integrate with OS accessibility infrastructure at all. That means any UI built with it will be a black box to anyone who needs accessibility tools like VoiceOver, text magnification, or semantic navigation.
It should be. There's AccessKit integration in egui (the most prominent Rust immediate mode GUI implementation) now, thanks to Matt Campbell's work. I would welcome GPUI and Zed adopting that, and I'm sure Matt can offer help in getting it integrated.
Zed developer here, I've actually looked at how to implement AccessKit in Zed, though I couldn't find a quick way of jumping into it beyond 'read the egui PR' at the time. If anyone wants to take a stab at it I'd love to help out and make it happen :)
Hi, lead AccessKit developer here. Sorry I haven't written much documentation yet. For now, the best way to learn how to use it is indeed to study the egui PR.
It is, Flutter (which also draws all the UI on a canvas) already does that - it’s called AccessibilityBridge. The way it works is they hook into the native accessibility system and create virtual accessibility nodes with the same size and coordinates where the “widgets” are drawn.
I think it’d be useful to create some common AccessibilityBridge-like library since more and more UI frameworks are taking the same approach as Flutter and GPUI.
we absolutely want to make accessibility work – hopefully open source will give us a bit more wiggle room to have folks help us figure some of it out as well.
I'm really happy the Atom team is back with another editor and atoning for the sin of creating Electron.
I'm glad Zed exists and hope it succeeds. I'm rooting for the team and hope they can put Microsoft's editor down the way Microsoft put Atom down.
With that being said-- I won't be using it. I've been using Vim for ~15 years and I've tried every editor and IDE under the sun. I just can't use anything else.
No, because a game utilizes the GPU 100% most of the time, while the simple GUI of this editor doesn't require as much calculations (fewer triangles, etc.). So your GPU can downclock and stay cool and quiet.
What if my GPU is totally off most of the time and it turns on my secondary GPU for games which increases the heat of my laptop considerably even doing very little? Do they let me select which GPU is used or just the 'default' because that default is a really bad choice for a terminal.
if you're a developer and dependent on mac you're probably building "the next big saas" in "the next big tech company" which will fail in few years.
a developer tool should be available on ALL platforms and specially Linux, if a tool is not available in linux no one will adapt it no matter how shiny it is.
"Ah yes Linux , the classic developer os"
it's not available on WINDOWS and on LINUX.
which is like THE MOST USED PLATFORMS ON THE PLANET.
You apple fanboys try to find excuses for everyone
Anecdotally, Mac has become very popular at my workplace, with a growing number of developers choosing Mac for their primary device. We mostly target Linux, so regardless of workstation we're usually remoting into a Linux box or running some VM/Docker setup locally. When I can work on projects locally without a VM, the interactions on Mac feel a lot more natural compared to Windows.
I don't think I would switch back to a Windows machine any time soon, although if I was offered a Linux laptop I'd be very tempted.
I don't really consider the SO survey useful anymore. It seems to be hyper-selected to just Silicon Valley techbros, instead of either the more general US tech industry, or the worldwide tech industry.
If the Silicon Valley is just a bubble, then the two bubbles I live in (life-long FOSS/Linux guy, but also deep into the hosting/transit/non-eyeball end of the Internet), every single Mac owner I know does not run OSX, but only wanted it for the physical build itself; they either run Windows because of job reasons, or they run Linux because they like to get shit done.
Everyone else I know is slowly joining #TeamFramework. Having made what seems to be the best Ryzen laptop on the market is certainly turning heads, and congrats for pulling it off.
Yes? I literally just said "HN hyper-selects their own bubble, but if I do it, I get entirely different results; neither of them are likely to be correct."
The reverse of this is, I don't know how SO can validate the claim, or the accuracy of any of the results in the SO survey.
If a lot of people self-select themselves to align with techbro culture, techbro culture is mainly in the Silicon Valley but not only there, and they think, for example, owning a Mac is how you signal belonging to that group (as in, Macs are a modern day Veblen good), then people who think they are (or wish they were) part of the techbro culture, no matter their location, will buy Macs.
So, the claim being given by people quoting the SO survey is: SO survey responders are not mainly made up of people who self-identify as techbros or have had their purchasing decisions (such as buying a Mac) greatly influenced by techbro culture.
I do not know of a way to argue that position. You can state that, literally, techbros live outside of the US, which is obviously true without quoting this survey; it however, does not mean the Silicon Valley doesn't act as an echo chamber, and doesn't have great effect on the tech industry worldwide.
What I would more likely want to see is, of that 21% of America, how many live in either the Silicon Valley or the Seattle/Redmond region. As in, how many are hailing from the Chicago or New York or Dallas metros, or non-metro regions entirely.
I'm guessing that, still, many would have responded that they live in either the SV or the northwest, greatly more-so than elsewhere in the US.
Yes, I run docker for a lot of workloads, and I can't drop the $3000+ they ask for a 32GB machine, where the same upgrade is like $100 for a regular Windows laptop.
There are obviously more programmers on Linux and Windows than on Mac, no matter your bubble. Since zed is Mac only, necessarily most programmers won't use it
Sure, but if you live in the bay, you'd be forgiven for assuming that 80% of developers use macs, 17% linux and 3% windows. Developing mac first makes perfect sense to me.
yes, the title of this post should be, "Zed, a collaborative Mac-only code editor....". I also find it kind of obnoxious that their front page says they are focused on making the world's best code editor without mentioning that it is tied to one platform. I don't think it can come close to being the best editor with such limitations.
Would it be possible to have option+click+drag create multiple cursors where you drag and not a selection with cursor at the end? Basically, the behaviour that Sublime Text has—Sublime Text's multiple cursors are, frankly, awesome.
I have found shift+option+drag to create a rectangle, but no way to create vertical lines of cursors through lines of text, and my favourite—drag down to the right of your code to put cursors at the end of every line. I use that all the time: scrape to the right hit ctrl+a to get to the start of a line, cut out and replace text, ctrl+e to the end, add more text etc. Really useful for turning data into code.
Option+click+drag is one of the most useful features of Sublime Text and my muscle memory is stuck to it — to the point where I keep Sublime Text around purely for the feature (well, this, the fantastic sort/permute lines options, and the live-highlight of regex searches that give you instant feedback of whether your regex is working correctly). Sublime Text is where I do all my data formatting, and where I test every regex before using it in my code.
One issue I've struggled with in Zed is full language support (linters/formatters). Has anyone at the Zed team thought about integrating a metalinter like Trunk Check (https://docs.trunk.io/check)?
The homepage has a benchmark that compares Zed's "insertion latency" to other editors, and this is the description:
> Open input.rs at the end of line 21 in rust-lang/regex. Type z 10 times, measure how long it takes for each z to display since hitting the z key.
Could someone clarify what that means? My interpretation of that was to go to https://github.com/rust-lang/regex/blob/master/regex-cli/arg... and start typing 'z' at the end of line 21, but that doesn't seem to make any sense. I guess that repo got refactored and those instructions are out of date?
They are opening the same specified code file (input.rs) in each editor and measuring event/process/highlight/etc pipeline duration.
Which is straightforward, but also depends a lot on how each editor is configured for Rust support and what functionality you are actually getting in that time.
No way can I use a text editor that requires the hi-perf discrete gfx card. My battery just can't take it. Hope I can get all my work done in about 1.5 hours, because that's how long the 2019 16" mpb will last with the discrete card running.
oh, sorry, you asked for an IDE and not an operating system... /s
seriously, budget a day or two for the ergonomics and a week for wrangling plugins, but some of the most productive developers in the world use emacs and they never need to worry about vendor issues, porting issues, not having a GUI, support for some weird file type, ability to create some funky type of macro, etc.
I tried to get into emacs a few years ago and just got fed up with the incredible regularity with which packages broke during routine updates. It was at least an hour a week on average, I feel, conducting bug hunts for either minor (this or that keybind/function doesn't work) or major (I can't get any packages to load) bug hunts. Was I doing something wrong? How do emacs people deal with this?
Admittedly, I was using Spacemacs which probably has way more packages than a bespoke emacs configuration, but you must understand if people get turned off when they ask you how to use an editor and your answer is "first, spend a few weeks setting it up and understand how each of your many packages works".
The same way anybody working with any package system does, by pinning ancient versions and/or just never updating all packages, at least for packages that aren't so widely used that they pretty much are never broken
That's an outrageously false statement. If it were true, you'd rarely need plugins.
> I wonder what you'd consider an IDE nowadays.
The definition of an IDE hasn't changed in at least 20 years. INTEGRATED development environment.
If I install a C++ IDE, I have everything I need out of the box, and the experience is (usually) consistent across all features.
If I want to do C++ in Vim, I need to install about 10 plugins (to begin with) [1], which will result in a disjointed experience (each plugin has different authors with different visions) where things break randomly and I don't know why. Speaking from experience, unfortunately.
Yes, you can make it work and you can get used to it, but it's just a text editor that you try to coax into doing what you want by using plugins.
Whereas the IDE will give you a language-specific tool out of the box, without any significant effort or inconvenience on your part. And the overall experience is better because the IDE "just works" most of the time.
I won't switch from neovim but I gave this a try and I quite like it. It really is very fast and snappy. If I were using VS Code I would've switched to this one.
Opinion: very lovely and has a great potential.
Feature request: toggle dark/light theme depending on system theme, just like VSCode and JetBrains IDEs.
I love this, they are doing things in the best way from my perspective. They've used the best copyleft licenses (GPL3 + AGPL3) and they are clear that free/open source software does not mean you cannot still make money. As they say, commoditize your compliment. I wish them success, and this move makes me want to look into this tool a lot more now.
This is really exciting. I currently use Sublime Text 4, and coupled with SublimeLSP, it does almost everything I need. What's missing is good debugger support and UI integration. What's the story like for Zed on that front? I briefly scanned the homepage and did not see mention of a debugger.
Hi,
Great article! Your insights are spot on [topic]. I especially appreciate your point about [specific topic]. It's evident you've done your research. Keep up the excellent work! Looking forward to reading more from you.
Emacs is great many things, but fast it is not. Nativecomp improved some stuff a bit and there has been other improvements lately too, but ultimately the core of emacs is built for flexibility, not for perf.
A lot of things, from limited keybinding support due to terminal limits, unergonomic extension language, very poor default setup, lacking good GUI, poor multicursor support, lack of core Zed's feature - collaboration
Those features are a worse substitute for that awesome generic UI mechanism called immediate visual feedback (for the same reason Helix's visual-first mode, which unfortunately neovim also lacks, is better)
* god damn this is fast... vscode feels so bloated by comparison.
* lsp integration works great. loaded up a rust project I have and it has good support for autocompletion
bad: what keeps me from using it as my daily driver for now
* no support for integrated debugger. I use rustrover for rust dev sepcifically because its so easy to set debugger stuff.
* daily workflow inolves entering a docker container or sshing into a remote server to edit code. I don't see anywhere in zed where I can do this.
bottom line: I see a lot of potential here. I'd love to use this when I'm just editing some rust code and I need zero distractions but the lack of ssh remote makes this a no-go for all of my paid work at the moment.
I remember listening to a changelog episode about zed ages ago. I liked what I heard about but that fact it's not for Linux just now and I don't have any friends to code with means I'd likely not use it.
> Extensively fuzz tested for stability
> Performance and power mean nothing without reliability. That's why we've subjected Zed's critical code paths to randomized tests that help us find and fix rare edge cases. By creating controlled chaos in development, we achieve stability in production.
How effective is this approach? Does it catch many issues that would not otherwise have been caught through a more risk-based approach? Conversely, does it tend to catch any significant issues?
At least these days we order the themes by dark then light... Themes used to be grouped by name, so you would get rainbow flashbanged moving down the list!
This is very nice, I hope we have a windows version soon. And this might encourage Sublime Text to do the same, it's one of the most responsive editor for now, let's see.
I was hoping that zed, because it was designed from the ground up to be a collaborative editor, would allow me to start a dedicated server (headless instance) on a development server/container, and then let me and my colleague both connect to it and work together.
Such functionality would be wildly useful for both remote development and pairing, dev containers etc.
Instead I'll have to stick with VS Code remote develop + live share, of which the latter feels like a monkey patched afterthought.
Why would you need to have "faith" in an open source product? You can always get an older copy or build it yourself, or it could be forked by the community if it's popular enough and the original gets completely abandoned.
For that matter, why does an open source product need to make money in the first place?
> Why would you need to have "faith" in an open source product?
Well, even ignoring the central servers which are required for all of Zed’s collaboration features, I’m not personally interested in maintaining my own persona fork of Zed
So that means I need to “have faith” that someone, whether a paid team of developers or an unpaid group of volunteers, will continue developing, or at least maintaining the project.
Or that it gets abandoned in a state that I can keep using it indefinitely, I guess.
> For that matter, why does an open source product need to make money in the first place?
That’s a fine question in general, but in the announcement they pretty clearly say that they are hoping to make money with unspecified future subscription services, not that they don’t need to make money
I’ve been trying out Zed for the past two days. It’s for sure way more zen and focused compared to VS Code and I really enjoy that. At the same time, I felt the syntax highlighting wasn’t to the degree I’m used to and so I had to look a little longer to find certain words I had typed. All in all, I’m trying to find ways to use it because of how smooth it feels
The thing about the Zed workflow that I don't understand is local servers. Say I were to develop a full-stack web app using some frontend meta-framework (e.g., Next.js) and a backend running locally for development (e.g., Supabase). How would that workflow work with Zed? I need that live preview/reload available to be productive. Some sort of feedback loop.
Nice that it’s installable Homebrew. It’s very fast comparing to other GUI editors. Happy to see Vim mode and would love to see more of Vim commands enabled. I can see that you can open splits, but you can’t specify what file to open in a split. But you can move the splits around with the same shortcut as Vim. Looks very promising!
I am not in a market for a new editor... not even a little bit.
But overwhelming positive experiences here, that I read, people being happy with using it made me try it out.
It really feels different and better. Can't say why and what it is. But first impression is very positive, even the theme, I didn't feel the need to change it.
I had an interesting thought reading their post - is there a brand new open-source monetization strategy in “operate the AI assistant for your tool”? Users can always plug in a local LLM but the fine-tuning and inference optimization is potentially much better from a paid team focusing on the problem.
This is an impressive piece of engineering, but I already have a collaborative code editor. It's called vim+tmux, and I can use it over hardware going all the way back to the 80s. The older I get, the more I feel that if I can't use it over a VT220, I don't want it.
Is there a way to explicitly disable Copilot/AI/collaboration features? I won't be able to use this at work if there's a risk that it'd upload my code. It looks like a very fast editor (with Vim keybindings!) and I'd love to use it even without AI :)
I don't see a way to disable all the collaboration features, but it does look pretty difficult to do it by mistake. You have to add collaborators explicitly.
Actually kind of interested in this... I think collaborative code editing is the future of development.
I think the next evolution is also combining the collaborative code editing with a virtualized environment. Where it builds and deploys remotely or runs the code for test and validation.
Looks awesome. Doesn't support syntax highlighting for Solidity yet and from a quick search, I couldn't tell if there's any plugin/extension I could install (or how to do it). Any pointers?
Does anybody know if I can add custom keybindings to switch between the code editor and the files tree? Also, can someone evaluate how Zen is configurable in terms of "only keyboard" usage way?
A really really solid editor with some very good choices made, with exactly one exception: The LLM integrations. It's annoying to disable (3+ settings in different ways?) and by default on.
This is awesome news. Thank you very much!
I looked at Zed some time ago and even though I liked it, it being closed source and not very popular didn’t make me “trust” that it was gonna last.
Interesting that they don't compare their speed to Vim, as while it's not an IDE per se, Vim is still widely used as a code editor. It'd be interesting to see the comparison.
would be nice to have the collaboration feature something like LSP with an easy to implement protocol. I can see that being useful but you gotta let the people choose their tools too. If I can run neovim/jetbrains/visualstudio and can connect to someones Zed Collab protocol beacon, that would be awesome af for at least the basic features (file browser and seeing live edits). This whole notes/chat thing can stay a specific zed feature IMO
I just tried it and it's one of my favourite text editors for MacOS already. Lite-XL and VS Code are my others. All other text editors for MacOS are too weird.
I have 2 Macs, PCs, and Linux. I need a cross-platform editor, for which I use emacs or VS Code right now. Will Zed eventually go cross-platform before 2025?
Yep, JetBrains is the gold standard IMO. If you want something that works well for most cases and you don't want to think about it, just go with JetBrains. Even their free stuff is high quality.
What do you mean? "L" and "H" do move the cursor right and left for me just fine. It does not have macro recording/playback or extended commands though so it's not a full fledged vim emulation mode unfortunately.
Yeah, it's pretty annoying for any long-time vim user to not be able to use the macros, which Zed doesn't support. Biggest issue I've had with this editor for sure.
This is my problem with all "vim modes"; the 90% of bindings that they cover works until your fingers' muscle memory wakes back up and tries `gqip` or some other niche vim thing.
They address this in the linked article, under the heading "Wait... Doesn't Cash Rule Everything Around Me?".
f.e.,
> Zed Channels is one example of such a service. It's free for anyone today, but we intend to begin charging for private use after a beta period of experimentation. Providing server-side compute to power AI features is another monetization scheme we're seeing getting traction.
Another FAQ explains that enterprise support is coming. So I think they will go with the free/OS model to saturate the market, but they will charge for corporate licenses. That's something investors can understand. Hopefully they didn't pick the greedy ones.
Not "most" but certainly many. I love remote dev (via Remote Containers in VSCode) for large C++ codebases (which could not even compile on my Macbook because of different architecture). Running `clangd` LSP on my local machine would be a nightmare (it could easily eat dozens of CPU cores and tens of gb of RAM when indexing), but on remote workstation it is a breeze. Also I could work on multiple branches simultaneously, spinning multiple VMs and running VSCode remotely on them. So basically my Mac is just a typewriter and everything heavy happens on a remote VM — isn't it nice?
I have never tried the remote ssh setup (even though vscodium is my primary editor). I don't know anyone who uses it daily, and I've never heard anyone irl mention doing it. I think you're slightly biased here by your experience, and assume everyone shares it.
Edit: hell, I've wanted to check remote ssh in vscodium now to know what I'm missing, and it looks like they basically contain MS DRM and are only compatible with official vscode. License is also proprietary[1] and makes it impossible to use it with vscodium or other OSS projects. Yikes.
Platform support is in their FAQ, but I agree it should be a bit clearer on the homepage. Perhaps the download button should say "Download for MacOS" or similar.
I do think your comment here is unnecessarily rude.
Rude to who exactly? OP criticized the landing page for a) omitting the fact that ~85% of desktop computer users can’t use Zed, and b) alluding to this not being the case with language about working anywhere. This isn’t an attack on the authors, or on the product they clearly put so much time into.
Yeah, OP said the page is really bad. But IMO we should be able to express negative opinions on things we think suck about technical content without being forced to couch our language in euphemisms for fear of being rude, especially on a site like HN.
> Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
> Please don't post shallow dismissals, especially of other people's work.
Expressing negative opinions is fine. But perhaps you should have a fear of being rude. We are a community, and there's plenty of other people reading these comments. I value kindness and being constructive, neither of which the comment I replied to appear to express.
If you value kindness, consider empathy. The person you're somewhat haughtily tone policing is obviously annoyed about wasting a totally unnecessary amount of time on something ultimately useless to them.
And the feedback was constructive (and easily actionable): be upfront about the very limited platform support, don't waste people's time.
You're trying to high horse a complaint about someone referring to a new open source release as "worse than bad"? Because it doesn't support their preferred operating system? And it took them too many clicks on the website to learn that? No, that won't do.
I don't think there's a way to talk yourself back up onto the horse here. Drive-by barbs like "worse than bad" are a plague on this site. It was a rude complaint about the website for an open source project, and it was called out civilly. Ironically, it's your personalizing response that skirts closest to incivility.
You can just disagree next time (if you really want to) without lecturing people on empathy? I think that strategy might work better.
You seem really choosy about whom you apply your empathy.
If I had chosen more direct words, and written my criticism with less emotion, then you clearly wouldn't have any issue with it.
My comment has 81 points at the time of writing this one. Clearly the majority opinion here is that the substance of it was worthwhile.
God forbid humans express emotion on the internet. You are welcome to criticize my expression. To simply filter it out as bad internet speak is, at the very least, a bad strategy.
You used a tropey put-down. You couldn't even find an original way to dunk on the project. I have a lot of empathy to allocate this scenario, but it's unclear why any of it should go to you.
I don't mean to keep a flagged, unproductive subthread going, but since you're listening to me: stop writing that way on HN. Stop dunking on people's open source projects. Be careful with how you write criticism. The preceding commenter was absolutely right to call out the snark in that comment. If you don't call that stuff out, it spreads like kudzu, and pretty soon we're all just writing Youtube comments. They did you a favor, and you should thank them for it.
It's rude to make misleading marketing statements as well, especially when you are addressing a technical market. That is pretty bad, and thinking so is not rude.
The "Work with code on any machine" is confusing because it's not what you expect by glancing over the headlines. In the section they describe what they actually meant by it... It means that you can collaborate with others who have zed installed and run things on their computer, edit code etc.
I vaguely remember them being on the Rust podcast and I believe they said that they want multiplatform support but they first need to nail it on one platform before they go for the next.
I'm frequently confused when coming across hyperbolic comments like these. You can just say: "feedback: make it more clear that it currently only supports MacOS". Calling it worse than bad is a stretch.
It is not a lie, it just doesn't mean exactly what you want it to, and is certainly unintentionally misleading. You can work with code remotely on any machine.
My reply was to contextualize that it wasn't a lie. Maybe a poor choice of words, but given the context on the page, it's clear what it is talking about, which isn't the application being cross platform. A lie implies deception, which I think given the context on the page there wasn't any intention to deceive, just poorly chosen words that can have different meanings depending on the context.
> Work with code on any machine
> When you join a teammate's project, you can navigate and edit as if the code is on your local machine. Open any file, type with low latency, and interact with language servers. It all works seamlessly, whether you're working with someone at the next desk or on a different continent.
I'm curious what part of that indicates that it means any operating system? It is clearly talking about treating remote machines as local machines. Yes, words have actual meanings but they also have context in which they are said or printed.
I'm assuming that chatgpt agrees with you, personally I'm avoiding using their software until they figure out a better relationship with content creators / artists than they have right now. I still think calling it a lie is hyperbolic and inflammatory when it's just mediocre copy. Hope you have a great day.
I'm expressing an opinion here. It applies as specific feedback, sure, but it was also intended for anyone casually listening: this is a very generalized problem that I would like people to be aware of.
This might be bad in its own way, but I saw that the screenshots were 100% macOS and immediately assumed that it was Mac only and closed the tab. I actually had to go back and look because I could have sworn that I saw it written somewhere, but no, it was just the screenshots and the general feel of the text that clued me in.
I think part of it is that apps that are designed to be multi-platform don't tend to use macOS screenshots because maxOS is a minority of their users.
Apple has an entirely unique GUI framework (Cocoa) on top of an entirely unique rendering stack (Metal).
I understand why they would choose to implement Zed using those tools, but the sacrifice is platform compatibility. I'm not bothered at all with the fact that this is MacOS exclusive. I just want to know before I go try to download it!
It’s likely that Zed does not want users, or wants specific kind of users as this stage of their company maturity.
Mac developer is a proxy for the user type Zed is probably aiming to sell service: Western, wealthy, senior. Whileas Windows users are everyone and their uncle. Scaling to large userbase does not make sense if they are not going to pay for your product.
There is! It's called Darling [1]. Unfortunately, its GUI support is experimental and still very limited, so you won't be able to run Zed right now, but I still think the project is very cool.
Adding Copilot or GPT-4 autocomplete support to an editor is not hard. You can find 100 different projects on GitHub that do this which are all just calling OpenAI. It is vastly different in scope and design than porting native applications.
In product management parlance, this is low hanging fruit.
Easy to implement and capitalizes on something the market is highly interested in at the moment.
Given the concentration of developers who use Macs, and the accelerating interest in LLM-assisted coding, this makes a lot of sense to me.
And while it may be true that AI is a craze, this is also a very real and useful feature grounded in a solid use case that has high impact on its users unlike some of the crazes of the past that were purely built on hype.
Maybe I am in a different bubble/country, but most professional developers and hobbyist hackers I know have either ThinkPads or Framework laptops running Linux or sometimes Windows, if they have to.
This is a pretty useless poll when trying to compare MacOS, Windows and Linux usage, because they split Linux up into different distributions. Because adding all numbers up, you get >100%, I suppose multiple-choice were allowed, which doesn't even allow adding all Linux-like operating systems together.
So not really a good Argument that MacOS is used more...
"Given the concentration of developers who use Macs"
fairly certain this is a silicon valley thing, the vast majority uses Windows for development and I wouldn't be surprised if macOS and linux are pretty even these days
I think you may be right. This comment was based on 90% of the people I interact with using Macs, but those interactions are indeed mostly in SV.
In that sense, the “concentration” is in the SV region, and may have been a bad choice of word by me when thinking about the broader ecosystem.
With that said, macOS enjoys ~33% share in the latest Stack Overflow developer survey [0], and I think this illustrates the point somewhat. This is a healthy enough portion of the market to focus on, especially in the case of implementing low hanging fruit.
You never built a product if this is your honest take.
You should have a focus when you first build a product. You shouldn't be holding up your product progress because you don't support every platform.
This is the reason why you see a lot of apps support iOS before they release an Android app.
It better to build a good working app for one platform than to not build one at all because you are spending all your time supporting as many platform as possible.
I’m a Windows user so I was sad about this but I think their reason for doing it like this makes a lot of sense. I’d do the same in their shoes, I think.
Tangent: I wish we could standardize real-time code sharing the same way we have with LSPs. It is sad that if you want to pair-program, both have to be using the same editor. It really makes more sense to have one standard format so that one person can use VSCode and another can use Vim or Zed. If you want to make it standard practice you have to enforce everyone to use the same editor, which sucks.
Already at work some people have complained that I don't use VSCode for code-sharing.
With Neovims headless capabilities, I am hoping for the day every major IDE and editor has first party support as a front-end for Neovim, and then any Neovim plugin just becomes universal across all IDEs and editors. Which... you can make Neovim plugins in almost any language, I've seen bindings for "lesser popular" languages.
I'm not even a vim guy, but if JetBrains finally adds front-end support for Neovim, I might just become a de-facto VIM guy. Instead of emulating vim like they do, they could just literally support Neovim as a back-end. I'm still surprised nobody at JetBrains has put in effort into this.
Yet another front-end I think needs it. I am looking at both Sublime and JetBrains because I use both. Notepad from Windows by Microsoft will support Neovim before both at this point (I was shocked to find that it supports tabs, and lets me keep unsaved files open just like ST).
Real time document editing and sync is actually a tall order. To make it editor/tool agnostic you'd need some kind of overlay fs + first/third party support in the various editors (as most editors don't hot reload file changes; they expect the files they have open not to be changing underfoot).
Yeah I don't know what work is involved to build something like it, but not having it def. leaves the editor ecosystem in a vulnerable position for monopolistic practices.
After almost a decade it never got off the ground. And while it might address the core problem (collaborative editing), that still leaves the ecosystem problem: fuse mounts, first/third party adoption in the editor or via extensions/plugins. Arguably that's drawing the rest of the f*** owl.
Xilem seems to be heavily focused on using WebGPU. One doubt I have is: is it possible to have tight integration with the system's compositor using such an abstraction? For example, on Windows looking at the documentation for the DirectComposition, it seems that things are designed precisely to work optimally with DirectX. So my impression is that one would really need to use the native graphics APIs on each system to achieve optimal input latency.
It's a good question and one we're looking into. We're not planning tight integration with the system compositor on our 2024 roadmap, partly because it's extremely platform dependent, and presents compatibility challenges (you basically don't get it on X). But even for damage regions, which is on the roadmap, we're going to need to extend wgpu.
As computers got more complex, latency increased. It's really pretty remarkable to see how things have changed over the years[1]. I do agree that 58ms seems far outside of the range of "next display refresh", however.
> These are tests of the latency between a keypress and the display of a character in a terminal
Wow, the numbers are entirely unexpected. I want to see the latency when you have no display server, having booted directly to a shell.
I imagine modern competitive twitch-reflex videogames (fps, fighting) must have far less input latency - 100ms must be totally unacceptable. Suppose that's the power of a GPU.
Same. I'd hope that using LSP and Copilot would not meaningfully impact insertion latency, though - I'd expect those to be fully async, in a separate thread.
Consider discarding the CLA if you anticipate third-party contributions. If you need to verify provenance, the Developer Certificate of Origin is a better approach. Otherwise, this (correctly) reads as a project which intends to take advantage of FOSS contributors for a while before pulling the rug and making it non-free again.
Also the "CLA" link just points to https://zed.dev/cla, which asks me to authorize Zed Industries to access my private GitHub profile. So at this point I don't even know what the CLA says. That's important because not all CLAs are created equal. I think some don't fully remove the contributors' ownership, which would prevent the hypothetical future closed-sourcing.
I took one for the team and authorized access for long enough to read the darn thing, and yes, it expects effectively full copyright assignment and leaves the door wide open for a rug pull.
Zed Contributor License and Feedback Agreement
Welcome to our Contributor License and Feedback Agreement! Here's a quick breakdown of what's inside:
You (the contributor) are entering into an agreement with Zed Industries, Inc.
You're giving Zed permission to use and share your contributions (like original works or modifications).
You assure us that the contributions are truly your own and you have the legal right to share them.
You're not required to support your contributions but you're welcome to if you wish.
If you ever notice an error or change in the details you've given us, you agree to let us know.
In short, this agreement covers the terms for your valuable contributions to Zed's projects.
Legally binding section follows:
By submitting a Contribution, use of the Solution or any components related thereto (as such is further defined in Zed’s End User License Agreement located here: https://zed.dev/eula) or as enabled pursuant to your use of Zed’s collaboration tools offered therein, you hereby accept and agree to the terms and conditions set forth in this Contributor License and Feedback Agreement (the “Agreement”) for Your present and future Contributions submitted to Zed Industries, Inc. (“Company”). Except for the license granted herein to Company and recipients of software distributed or made available by Company, You reserve all right, title, and interest in and to Your Contributions.
1. Definitions.
“Contributor”, "You", and "Your" shall mean the copyright owner or legal entity authorized by the copyright owner that is making this Agreement with Company. For legal entities, the entity making a Contribution and all other entities that control, are controlled by, or are under common control with that entity are considered to be a single Contributor. For the purposes of this definition, "control" means (i) the power, direct or indirect, to cause the direction or management of such entity, whether by contract or otherwise, or (ii) ownership of fifty percent (50%) or more of the outstanding shares, or (iii) beneficial ownership of such entity.
"Contribution" shall mean Feedback (as defined below), any original work of authorship, and any modifications or additions to an existing work, that is intentionally submitted by You to Company for inclusion in, or documentation of, any of the products owned or managed by Company (the "Work"). For the purposes of this definition, "submitted" means any form of electronic, verbal, or written communication sent to Company or its representatives, including but not limited to communication on electronic mailing lists, source code control systems, and issue tracking or collaboration systems that are managed by, or on behalf of, Company for the purpose of discussing and improving the Work, but excluding communication that is conspicuously marked or otherwise designated in writing by You as "Not a Contribution."
“Feedback” means suggestions, comments, improvements, software code / snippets or other information submitted to, shared with or otherwise made available to Zed or its contributors with respect to, or in connection with the use or interaction with, the Work, Zed Network Based Service or Editor technology (as defined by Zed and as further described within Zed’s End User License Agreement located at https://zed.dev/eula).
2. Grant of Copyright License. Subject to the terms and conditions of this Agreement, You hereby grant to Company, and to recipients of software distributed by Company related hereto, a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copyright license to reproduce, prepare derivative works of, publicly display, publicly perform, sublicense, and distribute, Your Contributions and such derivative works (the “Contributor License Grant”). Further, to the extent that You participate in any livestream or other collaborative feedback generating session offered by Company, you hereby consent to use of any content shared by you in connection therewith in accordance with the foregoing Contributor License Grant.
3. Grant of Patent License. Subject to the terms and conditions of this Agreement, You hereby grant to Company and to recipients of software distributed by Company a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Work, where such license applies only to those patent claims licensable by You that are necessarily infringed by Your Contribution(s) alone or by combination of Your Contribution(s) with the Work to which such Contribution(s) was submitted. If any entity institutes patent litigation against You or any other entity (including a cross-claim or counterclaim in a lawsuit) alleging that your Contribution, or the Work to which you have contributed, constitutes direct or contributory patent infringement, then any patent licenses granted to that entity under this Agreement for that Contribution or Work shall terminate as of the date such litigation is filed.
4. You represent that you are legally entitled to grant the above licenses, including but not limited to any video content shared or recorded as related to collaboration with Zed and the Zed community. If your employer(s) has rights to intellectual property that you create that includes your Contributions, you represent that you have received permission to make Contributions on behalf of that employer, that your employer has waived such rights for your Contributions to Company, or that your employer has executed a separate Corporate CLA with Company.
5. You represent that each of Your Contributions is Your original creation (see section 7 for submissions on behalf of others). You represent that Your Contribution submissions include complete details of any third-party license or other restriction (including, but not limited to, related patents and trademarks) of which you are personally aware and which are associated with any part of Your Contributions.
6. You are not expected to provide support for Your Contributions, except to the extent You desire to provide support. You may provide support for free, for a fee, or not at all. Unless required by applicable law or agreed to in writing, You provide Your Contributions on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation, any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A PARTICULAR PURPOSE.
7. Should You wish to submit work that is not Your original creation, You may submit it to Company separately from any Contribution, identifying the complete details of its source (e.g., attribution) and of any license or other restriction (including, but not limited to, related patents, trademarks, and license agreements) of which you are personally aware, and conspicuously marking the work prior to submitting it to Zed or the Zed community.
8. You agree to notify Company of any facts or circumstances of which you become aware that would make these representations inaccurate in any respect.
I like CLAs. If there's an open source project that I love/depend on, I will most likely want to be able to contribute to it for my own personal reasons at some point (e.g. fix a bug that bothers me or add a feature I want).
I am a Sublime Text addict, but hate how slow development is and how resistant they are to adding more dev-focused features. I would happily sign a CLA if it meant I could modify the code myself, even if I end up paying to use my own code.
Giving the owner permission to sell my contribution (which is likely very minor compared to the rest of the project) gives me added peace of mind since I know that the project is (or can be) sustainable.
A rug pull is always bad, but it's not fair to assume that that's what will happen (unless you know something we don't).
And besides, even if it does happen, that's what forks are for. This project is licensed with GPL and AGPL.
You do not need to give anyone special permission to sell your contribution, everyone in the world is equally entitled to make commercial use of (including outright sale of) free and open source software; without this it is not free or open source.
The only purpose of a CLA like this is to provide for a future where the software can be made non-free again.
What a weird statement. It's beautiful! You don't need comments, just read the code. No need to litter code with ambiguous language. The only place a comment fits is if the purpose is not clear from the code.
I dont think is much about how much you write, is more like your brain consciously and subconsciously keep track of how enjoyable/frustrating is to any use any given tool (or perform any given action), for developers that includes the code editor, and some developers percieve speed as one of the most important features to make it enjoyable
I think it’s more that slowness makes things less enjoyable. Snappy should be the baseline. I love IntelliJ, but the fact that it often just decides to randomly re-index, slowing the whole system to a crawl, makes the experience of using it extremely frustrating.
If it's open source, why do I have to agree to a bunch of terms and conditions that look suspiciously not like a standard open-source software license when I try to install it?
- It really is remarkably responsive,and makes one really notice how UNresponsive everything else is. I have reasonably fast machines, so we're not talking about the difference between 5ms typing lag and 500ms, but it's still pretty surprising. VSCode never felt slow on my macs until I started using Zed.
- They seem reasonably responsive to feedback. There was some contention around how search/replace was initially implemented, and the current builds have something much more usable IMO. I'm not sure how much that was driven by community feedback, but the changes were great.
- The debug syntax tree mode is a really neat feature that I think demonstrates how much more advanced zed is under the hood than older editors that are doing syntax highlighting via regex.
There are a few downsides that I'm hoping get addressed soon:
- The collaboration workflow/security isn't very clear to me. You sign in via github (no other option???), there are 'contacts' (I guess these are github usernames?), and 'channels' (where do these live? on zed's servers?). I would really like to know if I can self-host the chat server and use a company oauth provider rather than github. If the diffs being passed around are going through zed's servers, that may be a showstopper for the company I work for as well. If they're p2p and encrypted, maybe not.
- I would love to see ollama integration. This + continue is the only reason why I spend any amount of time in vscode now. There's an issue for it here: https://github.com/zed-industries/zed/issues/4424