FWIW I use GoLand w/ Supermaven, currently.
PS: One UI feature that I haven't seen anywhere yet is not splitting the screen into code and chat but unifying it.
Also a common approach (e.g. in openai composer) seems to be to modify the file line by line which takes a very long time if the file is long, which (agentic) tools use a diff approach instead?
Autocompletiok is losing features every day, it seems. Yes, Jetbrains might either be removing features to move them to AI, or breaking features involuntarily in order to move faster to implement AI.
Either way, AI is killing Jetbrains.
aider
Copy and paste workflow is a minor slowdown, but nothing compared to things like smart links in terminal, auto detection of run configurations, etc, etc.
I watch people navigate code in VSCode and I want to pull my hair out. Things that I don’t even think about are hard and/or require just falling back to search.
And before “there is a plugin for that”, I’m sure there is. I’m sure you can configure VSCode to be just as powerful as IDEA but the rank and file using it aren’t doing that work to install and configure a bunch of plugins. So, on average, VSCode doesn’t hold a candle to an IDEA.
With Aider I skip a lot of the copy/pasting but I’d still copy/paste to the browser before I left IDEA.
For me it's the other way around, when I see someone using an IDE instead of a lean editor I see their struggle. Multiple seconds to open the IDE (sometimes tens of second), multi-hundred millisecond lag when opening a file, noticeable input lag. And when you have to edit a file your IDE doesn't understand, all you have is a bloated notepad.
I know I'm biased and I intentionally wrote this one-sided to counter your post. In practice, it just depends. Right now in my work I primarily edit scripts (up to a few hundred lines of code), do quick edits to various larger projects - sometimes few different projects a day - and read files in dozens of programming language (I only reall program in Python and C/C++, but I have to constantly consult various weird pieces of code). VsCode works great for me.
On the other hand, long time ago when I was working on large C# projects, I can't imagine not using Visual Studio (or Rider nowadays I guess).
+1 this is what brought me back to vscode after experimenting with goland. To me vscode better handles the heterogeneity of my daily work. In my workspace I can keep open: a golang codebase, a massive codebase consisting of yaml config files, a filesystem from a remote ssh connection, a directory of personal markdown notes, and directories of debug logs. In my experience jetbrains excelled at the single use case, but vscode won on its diversity.
I will say that the parent comment had me curious about goland again. But I suspect I really need to spend more time configuring my vscode setup. I spent years using emacs, and would love to have a helm-like navigation experience.
I'm a near-exclusive user of VSCode (or Codium, at home) and like to think of myself as moderately advanced. I continually update my configurations and plugins to make my workflow easier and often see my peers stumble on operations that are effortless for me. It's hard to explain to them what they're missing until they watch me code. So now I'm curious about watching some typical Jetbrains workflows.
For most of the rough edges, I've found workarounds at this point.
I miss some of the refactorings, but half the time the AI does them for me anyways.
Smart links in terminal are supported. Detection of run configurations is supported.
My main issues atm are:
- Quick search is inferior. You can't search for folders, and symbol search is global, without the option to exclude certain locations (such as build artifacts) from indexing.
- cspell is more annoying than it is useful. I don't want to babysit my spellchecker. Without extensive configuration, there are far too many false positives.
Sublime Text's text search is the killer feature for me. CTRL+SHIFT+F and I can search a million LOC+ codebase instantly, navigate the results with CTRL+R, do a sub-search of the results with CTRL+F, set bookmarks with CTRL+F2 (and jump to them with F2/SHIFT+F2), pop the results out to a different window for reference, etc. And all that happens with no jank whatsoever.
The LSP plugins make life easier, but even without that Sublime is often able to find where symbols are defined in a project using just the contextual information from the syntax highlighter.
I tried CLion for a while, but couldnt get productive in it. Ofc I'm much more experienced with Sublime, so maybe I just didnt give myself enough time to learn it, but CLion felt sluggish and inefficient. The smart code features are probably more advanced than Sublime's LSP plugins, but I didn't find anything that would make the switch actually an improvement to me.
* Click to find usage is exceptionally good.
* when refactoring, it will also find associated files/classes/models and ask you want to change them as well. It’s also smart enough to avoid refactoring things like database migrations
* Click-run just about anything is amazing. I work in multiple languages in multiple code bases. I get tired of figuring out how to install packages, run things, and even get into debug mode. Most major tooling is supported out of the box.
* Debugging. Lots of data types have smart introspection when debugging. It knows that Pandas tables should be opened in a spreadsheet, JSON should be automatically formatted, and you really just want that one property in a class already formatted as a string.
* Built in profilers and code coverage. This is something that’s always annoying to switch tooling with.
* Great Git integration, though that’s probably par for the course.
* Database integration in some toolsets (like Rails). If you need to look at records in the database, it will jump you directly to the table you need with Excel-style filters
* Local history on just about everything. This has saved my butt so many times. You can delete an entire folder and know you can restore it, even if you delete it from Git.
* Automatic dependency change detection. For example, after a pull, it will identify if new or updated updated dependencies were pulled. 1-click to install.
* Type hinting in in-type languages
I have a laptop I bought 10 years ago. It only has 16 gig of RAM [1].
I have had 8+ editors open, mix of visual studio and vs codes. And in VS you often group all your codebases into single solutions. So usually have multiple windows from multiple projects open in each ide.
It only struggles when I leave the debuggers running several days because of a slight (known) memory leak in visual studio. There's probably a fix but reopening the ide takes like 10 seconds. And it remembers all my open files.
All editors are much better, faster at searching, use less memory, etc. than they wwre 10/20 years ago.
Everyone's improved. You seem to be a bit stuck with an old impression.
[1] I have a vastly more powerful machine but I keep procrastinating switching my work setup over to it.
Notice that if you work with large projects it is crucial to give the IDE enough RAM so it doesn't thrash a lot. You can also remove a lot of its default plugins to make it much faster.
Features I'd love:
* Reproduce the web/browser chat UI in the IDE, this is an easy concept to interact with vs a dialog that goes away after each run
* Provide tabs for multiple chats (each chat can have different files/history/etc)
* Allow multiple Aider processes running at the same time
But this is still super slick as-is, thank you!
Seriously, this here right now is the precise moment in time where people will either look back at wondering how such a clear leader managed to sink into insignificance, or not.
I love IntelliJ more than my own kids, but if they don't add "the AI does not just talk about what code to create, it actually creates it, across multiple files and folders", then I'm out.
Just yesterday I made Cursor rewrite the whole ui layer of my app from light-mode-only to light-and-dark-mode-with-switcher in one single sweep, in less than 5 minutes (it would have taken me hours, if not days to do it manually), and this is just not feasible if you have to manually copy-and-paste whatever Jetbrains AI spits out.
Jetbrains — Move. Now!
Claude 3.5 + Cursor has fundamentally improved my productivity. It's worth the 20 dollars a month.
I've written thousands of lines of Vitest tests with it, and they have come out near perfect. It would have taken me days to write those tests by hand, but now I can just generate them and review each to make sure it works.
Intellij will have it's lunch eaten if it doesn't pursue the Cursor/Windsurf editing modality.
I keep them both open on the same project, as there are some things IntelliJ does superbly.
I evaluated Jetbrains AI and Copilot with VSCode, but they just didn't impress me. I tried Cursor, and subscribed a couple of days into the trial. The workflow is just right.
If your main issue is the keybinding though there is a vscode plugin[1] that recreates Intellij IDEA bindings, which I found helped smooth the transition during my tryouts for me.
[1] https://marketplace.visualstudio.com/items?itemName=k--kato....
Personally, I'm using Goland for editing, along with Zed for its AI assistant. I just have the same project open in both, and do any AI editing in Zed.
I really like the AI UX in Zed with the chat window on the right being context for inline assist (and being able to easily include tons of files in the context window).
However I think JetBrains should be worrying. I've been using and paying for JetBrains products for many years and Cursor is the only thing making me consider switching from it.
We are allowed Copilot and Chatgpt because of an enterprise contract with each. Luckily copilot has been improving but there definitely was a while where it felt way behind the jumps other products have had in the past year or so.
A product can work however it works, good or bad, but if you can't wrap contractual guarantees around it that are palatable to your enterprise customers, you're not going to get enterprise sales.
GitHub Copilot works fine with these; just authenticate against github.com first
I would understand this mindset when it comes to consumer uses of it, but enterprise is where Microsoft makes its money and it would have to be the dumbest business decision ever to ruin the enterprise cash cow by doing that.
There are areas where Microsoft knows (or at least behaves as though) they have an effective unchallenged monopoly, and off boarding is too costly an endeavour.
Please explain. Vague insinuations mean nothing.
I was wondering if you have used Copilot Edits and can compare it to Cursor Composer? Is Cursor much more superior when it comes to multi-file edits?
Do you also have tips for how to give Cursor some documentation, is there a way to make it RAG on a folder of markdown documentation files?
For us who prefer coding ourselves (instead of telling LLMs to do stuff and review the result) it is much much better.
func(a, b, 1)
func(b, c, 1)
You just can't make Copilot do this. Even after you move to the second line, it won't suggest anything. It just suggests a completion starting from where your cursor is, instead of an inline edit around where you are at.
Sometimes you can delete everything after func( on the second line and Copilot will finish it, but sometimes it just can't and decided to autocomplete irrelevant (e.g. func(1, e, f).
In this case there's not much intelligence needed, but for more complicated changes Cursor does just as well.
Getting answers (with references) to questions *destroys* using Google.
Cursor has that, but also other modalities. It also just has a better UX for the core shared experience. Cursor also has their own models they mix with APIs to big well known models.
For example…
1. Cursor’s chat window generates patches (multi line, multi file, non contiguous) that can be directly applied to the file editor with a click, instead of requiring you to copy/paste the chat results. It lets you apply parts of the patches too.
2. Cursor supports multi-line changes (which show up as auto-complete), where the lines aren’t contiguous. For example, if you rename a field in a class, the AI will propose a rename for the Getter/Setter methods, as well as all the uses of said field/method. It’s like all the familiar refactoring tools available in IDEs, but AI powered so it operates “fuzzy matching”
3. Building off 2, they will use AI to move your cursor around (it’s not intrusive).
4. Cursor supports BYO keys for OpenAI, Gemini, Anthropic, etc. They host their own model which you’d need to pay to access though.
5. They support AI autocomplete and conversation in the command line. Helpful for remembering commands or if you need to change a command to test some change you’d made.
There might be more but this is what sticks out.