Their ACP integration is good (finally), at least with Codex. For when I want to be near the code, I prefer it now over TUIs or the Codex app. It feels much more immediate that way. Sticking to the basics will kill them IMO. They have been pitching themselves as agent-native for a while btw.
And the workflow they are describing in the post sounds amazing, I‘ve been wishing for that. It‘s orthogonal to CVS, no? You can still make PRs. But, crucially, you can loop people in early on, when it matters the most.
> Their ACP integration is good (finally), at least with Codex.
you must not use it much. try pasting in an image, quitting zed and restarting (ie: upgrading the app). it shows you the base64 encoded data instead of just showing the image. heck, it doesn't even show you the image when you paste it in. it just shows the filename.
ACP integration is very good, IMO. I'm using it with Claude Code. They've made so many improvements over time. I use image pasting myself, TBH I was excited it worked at all. Surprised that some rough edges around images are enough to ruin the whole thing for you.
images were just one glaring example... i didn't say anything about ruined, i just think a lot of those 'rough edges' are a bit more rough than they could be.
it is a code editor after all. how about integrating the LSP so that it doesn't have to resort to rg/grep/sed/scripts to edit code?
I agree with you that I wish they'd just nail the editor. I've been using it for years now and I really like it, but there are a few rough parts (settings and LSPs especially) that could probably be tuned up in no time if the focus was there.
Having said that, I love the idea of correlating the agent sessions to the changes. I'm glad someone is working on it.
You don't need a tool for that just configure your agent to write commit messages that link to a LLM log. And changes between commits. Maybe don't care why do we even care?
Did you even read the post? The idea is to tie conversations to code. This is a more sophisticated version of something like doltdb, not a git replacement.
Why develop this when the core editor is almost unusable? On Linux wayland, copypaste is broken, the file manager is laggy and snippets do not trigger before the heavy hint logic is done. This is just on top of my head, the list is long
"Almost unusable"? Slight hyperbole don't you think? I initially tried the Flatpak version on Fedora 44 and had a few minor problems, but since switching to the terra repos version (terrapkg.com) I've had zero issues, Have you tried that?
For me, its a massive improvement over IntelliJ for my daily workflow.
I can't use an editor where the file contents I'm being shown don't match the reality of the file. I guess this issue doesn't affect most people somehow, but in the age of CLI agents I'm not sure how more people aren't running into it.
I was really enjoying Zed for a short while before I ran into this fundamental issue.
All I can say is it's the only editor I've ever used with this type of issue and unfortunately it's a deal breaker for me. I can get by with any other editor even if they aren't my favorite, but only Zed made me realize that the most important thing an editor can do for me is accurately show me files.
Just installed it, works on my kde and niri setups (I logged out and switched to each), copy paste works as well. File manager laggi-ness I didn't notice I don't use it, I use it mostly to type stuff out.
I have never used snippets so unsure about that too, going from command bar seemed to work.
I think it might be bad, but could also be it's not correctly configured maybe try asking codex/claude or something to check if it can fix/fine tune the config, if that doesn't work file issues?
I do suspect it might be working for me cause I got lucky I do see an open issue[1], but given Zed is used a bunch the lack of comments does suggest maybe it's not that wide spread either way not a good look.
There are a lot of subtle bugs. I'm on macOS too, but perhaps my workflow is unusual? Every time I open it, I find small bugs. There are so many features that it's hard to maintain them, I guess.
For example, a lot of disable buttons do not actually disable stuff. They just hide it from the UI.
I started using Zed a couple weeks ago and immediately had issues with copy/paste/buffer on an M1 Mac. I tried a few suggested fixes and they didn't work.
It still doesn't actually display a window my end until I press a keyboard key or move the mouse. So very often I sit and wait for it to open, only to realize I have to press a random key for it to unhide.
I really love Zed and I want them to really succeed. Unfortunately it seems that they are swamped with various things and the issue tracker is not a priority for the team, which fair enough. OSS is a gift and I take it as such. Wild proposal but if someone is willing to start a "user advocate" contributor, I'll be happy to donate monthly so user requests can be taken a look at :)
I really do like Zed, it's snappier than VS code and with pretty good vim emulation. Even through SSH, the native UI allows for no network latency, compared to something like vim+terminal. Like many other posters, I do feel like the core editor is getting worse.
Some issues I've encountered recently at work:
- Routinely encounter bugs with the python language server, it's particularly slow when compared to using VSCode. All that snappiest disappears when finding symbols after the '.' is considerably slower than my typing speed.
- I found similar issues with C++ code bases.
- It's also so hard to debug when something goes wrong. I was trying to find out why pyright keep spinning on startup, but the "LSP logs" never show anything.
- Connecting via SSH always opens the repo in restricted mode, even though I check the "Trust all projects in" checkbox. Never happened before.
- Debugger UI is a mess. Make a run, stop the run, edit the task.json, try running again, and 2 tasks show up. One for the old and one for the new.
--
I stuck with it because I really do like the UI, the performance, and initially it seemed to be an improvement from VS Code for my workflows, but lately I've been moving back to code as it seems more stable. I don't really use agent features, so I don't think I'm the target customer for this or other priority Zed features.
I do see the usefulness of something like this, I'm interested to see if the new agentic style of programming will result in the "code" part of editors becoming second class (for the better) and will allow for more rapid development. Although I'm not looking forward to read my coworker's "threads" of how a feature was made, I have a feeling that it will be more difficult to read than the raw code, as from my experience, LLMs don't really employ terse and effective communication (might not be entirely fair, I found ChatGPT /study mode pretty good.)
--
Anyone know of a good replacement editor that might meet my usecase? I'm really a batteries included type of developer.
Looks like a great feature/product/functionality. I hope that Zed editor will make it work with LLMs/agents that run in the terminal, not just the Zen Agent itself. Is it even possible? I use OMP.sh agent running in Zed terminal.
This is the killer Zed feature, IMO, that they support almost any agent via extensions. I deleted Cursor but still use the agent via Zed because it’s a lot nicer experience, Zed never had popups telling me Grok is half off. I also use OpenRouter and Zed.dev with it, that’s three providers out of dozens that it supports.
Exactly. The more competitive the space is, the more important it is that products purposefully support 3rd-party providers. Now that people are increasingly switching from Anthropic/OpenAI for inference reseellers and on-prem compute, transparency and portability will be the new values, not being slightly better than the next agent or model.
I had the great pleasure/misfortune of being forced to work with LaminDB, and I can tell you folks; data provenance isn't going away -- it will be used to train the very AI models snapping at our heels
This feels very useful for training agents but almost completely useless to humans? Do I really want to go back to every change I made? Whatuse would that be?
i think a system that uses git and file system artifacts (txt/md/xml files whatever preference) is much better. eng teams need to define their artifacts, eg plan file, reqs etc whatever is needed and important to that team. the agent convo itself doesn’t add much value IMO and adds unnecessary bloat.
this looks like jetbrains "local changes" but globally synced. "for the agents" mights scare people who jump to think about micromanagers, but these people mostly hide or fake work. even micromanagers will eventually give you an agent to use so it won't matter what is in the change log, what matters is speed+quality. and since you can iterate from any branch you have inginite try and retry possibilities
Maybe I’m pessimistic, but I think this just gives micro-managers the data they need to micro-manage you. It invites people to scrutinize the process, instead of the results.
Not once have I wished to review a colleague’s conversation with an AI. I want to review finished work.
I look at this and I can only imagine layoffs being justified with “bad prompt quality”. An incident retrospective broadcasting your agent conversations to the entire company, with people speculating where you should have put the “make no mistakes”.
It’s a shame, really, because I like Zed as an editor, but I can’t bring myself to support them if this is the direction they’re choosing.
The sort of company that would do those things is going to be mandating you to use their enterprise AI plan anyway, which already lets them read all your conversations if they want to.
Letting them read my conversations is okay. Systematically reconstructing a timeline of every keystroke that went into a project, then immortalizing this timeline into a VCS, that’s the creepy part.
I suppose companies could do this anyway just by installing spyware in your work computer, but I would very much prefer if that spyware was not my editor of choice.
The way I’m seeing AI used by organizations is to produce process slop: tickets, summaries, docs, etc. rather than drive process change and optimization, or enable individuals to be more effective within the organization.
Jujutsu already does this. I had agent screwing up changeset over couple hour long session and it took one shell command to split on pieces with longer pause.
The only downside is that because of Jujutsu snapshots it can't be used inside sandbox (since all jj commands are mutating)
Snapshots have an intention behind them as a chunk that the programmer thought of. It has infinite flexibility depending on the programmer/team. IMO they are a necessary feature, but the article considers it a bug. It’s an explicit and deliberate chunking of code writing process which is what you want in a chaotic “collaborative” environment. The system doesn’t get to decide but you decide when to commit.
Yes, it is open source, and I am not saying they should not do it. Chromium is also open source, but if they added too much nonsensical AI functionality, I would be upset. I am free to improve it on my own, but that does not mean I cannot express how I feel.
I have not gone to the Zed repository and created an issue telling them to stop developing DeltaDB because it takes away the maintainers’ time.
https://github.com/zed-industries/zed/discussions/54150 failing to show newly created files and declining to provide a refresh button, instead adding a polling backend, breaks Zed on WSL.
Why a new version control system? Why not git, jj, or another existing system?
And the workflow they are describing in the post sounds amazing, I‘ve been wishing for that. It‘s orthogonal to CVS, no? You can still make PRs. But, crucially, you can loop people in early on, when it matters the most.
you must not use it much. try pasting in an image, quitting zed and restarting (ie: upgrading the app). it shows you the base64 encoded data instead of just showing the image. heck, it doesn't even show you the image when you paste it in. it just shows the filename.
it is a code editor after all. how about integrating the LSP so that it doesn't have to resort to rg/grep/sed/scripts to edit code?
Having said that, I love the idea of correlating the agent sessions to the changes. I'm glad someone is working on it.
For me, its a massive improvement over IntelliJ for my daily workflow.
I can't use an editor where the file contents I'm being shown don't match the reality of the file. I guess this issue doesn't affect most people somehow, but in the age of CLI agents I'm not sure how more people aren't running into it.
I was really enjoying Zed for a short while before I ran into this fundamental issue.
I have never used snippets so unsure about that too, going from command bar seemed to work.
I think it might be bad, but could also be it's not correctly configured maybe try asking codex/claude or something to check if it can fix/fine tune the config, if that doesn't work file issues?
I do suspect it might be working for me cause I got lucky I do see an open issue[1], but given Zed is used a bunch the lack of comments does suggest maybe it's not that wide spread either way not a good look.
[1]: https://github.com/zed-industries/zed/issues/47004
Maybe I got lucky?? Does it not auto update on linux maybe I am on an old build... Very suspicious.
For example, a lot of disable buttons do not actually disable stuff. They just hide it from the UI.
Terrible Linux support.
Shit, as someone who uses Zed everyday, I didn't realize what I am using is unusable.
- Vertical activity bar (https://github.com/zed-industries/zed/pull/58868#issuecommen... - rejected because it "doesn’t match our design vision for Zed")
This is my #1 annoyance: clicking microscopic icons at the very bottom of the screen to switch views.
- Don't crash my entire PC when there's a large (100+ MB) JSON file in the working directory
- Don't freeze when resuming from suspension (I suspect this is GPU-driver related on linux) forcing me to `killall zed-editor`
Some issues I've encountered recently at work:
- Routinely encounter bugs with the python language server, it's particularly slow when compared to using VSCode. All that snappiest disappears when finding symbols after the '.' is considerably slower than my typing speed.
- I found similar issues with C++ code bases.
- It's also so hard to debug when something goes wrong. I was trying to find out why pyright keep spinning on startup, but the "LSP logs" never show anything.
- Connecting via SSH always opens the repo in restricted mode, even though I check the "Trust all projects in" checkbox. Never happened before.
- Debugger UI is a mess. Make a run, stop the run, edit the task.json, try running again, and 2 tasks show up. One for the old and one for the new.
--
I stuck with it because I really do like the UI, the performance, and initially it seemed to be an improvement from VS Code for my workflows, but lately I've been moving back to code as it seems more stable. I don't really use agent features, so I don't think I'm the target customer for this or other priority Zed features.
I do see the usefulness of something like this, I'm interested to see if the new agentic style of programming will result in the "code" part of editors becoming second class (for the better) and will allow for more rapid development. Although I'm not looking forward to read my coworker's "threads" of how a feature was made, I have a feeling that it will be more difficult to read than the raw code, as from my experience, LLMs don't really employ terse and effective communication (might not be entirely fair, I found ChatGPT /study mode pretty good.)
--
Anyone know of a good replacement editor that might meet my usecase? I'm really a batteries included type of developer.
Looks like omp supports ACP, so all you have to do is specify a Custom Agent in Zed and it should just work.
https://omp.sh/docs/acp
https://zed.dev/docs/ai/external-agents
Is it available now? Still seems to be a waitlist?
Maybe I’m pessimistic, but I think this just gives micro-managers the data they need to micro-manage you. It invites people to scrutinize the process, instead of the results.
Not once have I wished to review a colleague’s conversation with an AI. I want to review finished work.
I look at this and I can only imagine layoffs being justified with “bad prompt quality”. An incident retrospective broadcasting your agent conversations to the entire company, with people speculating where you should have put the “make no mistakes”.
It’s a shame, really, because I like Zed as an editor, but I can’t bring myself to support them if this is the direction they’re choosing.
I suppose companies could do this anyway just by installing spyware in your work computer, but I would very much prefer if that spyware was not my editor of choice.
As anyone who's ever tried to split a days worth of work apart with `git add -p` is well aware.
It may be feasible with a million agents, but they could remember what they typed three hours ago anyhow.
Jujutsu already does this. I had agent screwing up changeset over couple hour long session and it took one shell command to split on pieces with longer pause.
The only downside is that because of Jujutsu snapshots it can't be used inside sandbox (since all jj commands are mutating)
See https://zed.dev/blog/introducing-deltadb but also https://zed.dev/blog/crdts from years ago.
I have not gone to the Zed repository and created an issue telling them to stop developing DeltaDB because it takes away the maintainers’ time.