To talk through the features, we engineered the following:
- A proactive agent
Agent which iterates on the linter errors (powered by the Language Server) and pulls in relevant context by doing go-to-definitions, go-to-references etc and propose fixes or ask for more files which might be missing in the context.
- Developer control
We encourage you to do edits on top of your coding sessions. To enable this, we built a VSCode native rollback feature which gets rid of all the edits made by the agent in a single click if there were mistakes, without messing up your changes from before.
- A combined chat+edit flow which you can use to brainstorm and edit
You can brainstorm a problem in chat by @’ting the files and then jump into edits (which can happen across multiple files) or go from a smaller set of edits and discuss the side-effects of it
- Inline editing widget
We took inspiration from the macos spotlight widget and created a similar one inside the editor, you can highlight part of the code, do Cmd+K and just give your instructions freely
- Local running AI brain
We ship a binary called sidecar which takes care of talking to the LLM providers, preparing the prompts and using the editor for the LLM. All of this is local first and you get full control over the prompts/responses without anything leaking to our end (unless you choose to use your subscription and share the data with us)
We spent the last 15 months learning about the internals of VSCode (its a non-trivial codebase) and also powering up our AI game, the framework is also at the top of swebench-lite with 43% score. On top of this, since the whole AI side of the logic runs locally on your machine you have complete control over the data, from the prompt to the responses and you can use your own API Keys as well (can be any LLM provider) and talk to them directly.
There’s still a whole lot to build and we are at 1% of the journey. Right now the editor feels robust and does not break on any of the flows which we aimed to solve for.
Let us know if there’s anything else you would like to see us build. We also want to empower extensibility and work together with the community to build the next set of features and set a new milestone of AI native editors.
Having a deeper integration with the editor allows for some really nice paradigms: - Rollbacks feel more native, in the sense that I do not loose my undo or redo stack - cmd+k is more in line with what you would expect with a floating widget for input instead of it being shown at the very top of your screen which is the case with any extension for now.
Going further, the changes which Microsoft are making to enable copilot editing features are only open to "copilot-chat" and no other extension (fair game for Microsoft IMHO) So keeping these things in mind, we designed the architecture in a way that we can go towards any interface (editor/extension). We did put energy into making this work deeply with the VSCode ecosystem of APIs and also added our own.
If the editor does not work to our benefit, we will take a call on moving to a different interface and thats where an extension or cloud based solution might also make sense
I want to get some broader feedback before completely switching my workflow to Aide or Cursor.
I’m still using Copilot in VS Code every day. I recently switched from OpenAI to Claude for the browser-based chat stuff and I really like it. The UI for coding assistance in Claude is excellent. Very well thought out.
Claude also has a nice feature called Projects where you can upload a bunch of stuff to build context which is great - so for instance if you are doing an API integration you can dump all the API docs into the project and then every chat you have has that context available.
As with all the AI tools you have to be quite careful. I do find that errors slip into my code more easily when I am not writing it all myself. Reading (or worse, skimming) source code is just different than writing it. However, between type safety and unit testing, I find I get rid of the bugs pretty quickly and overall my productivity is multiples of what it was before.
I can't tell if it is a UX thing or if it also doesn't suit my mental model.
I religiously use Copilot, and then paste stuff into Claude or ChatGPT (both pro) when needed.
The AI integrations are a huge productivity boost. There is a substantial difference in the quality of the AI suggestions between using Claude on the side, and having Claude be deeply integrated in the codebase.
I think I accepted about 60-70% of the suggestions Cursor provided.
Some highlights of Cursor:
- Wrote about 80% of a Vite plugin for consolidating articles in my blog (built on remix.run)
- Wrote a Github Action for automated deployments. Using Cursor to write automation scripts is a tangible productivity boost.
- Made meaningful alterations to a libpg_query fork that allowed it to be cross-compiled to iOS. I have very little experience with C compilation, it would have taken me a substantially long time to figure this out.
There are some downsides to using Cursor though:
- Cursor can get too eager with its suggestions, and I'm not seeing any easy way to temporarily or conditionally turn them off. This was especially bad when I was writing blog posts.
- Cursor does really well with Bash and Typescript, but does not work very well with Kotlin or Swift.
- This is a personal thing, but I'm still not used to some of the shortcuts that Cursor uses (Cursor is built on top of VSCode).
What features of cursor were the most compelling to you? I know their autocomplete experience is elite but wondering if there are other features which you use often!
The chat is okay, the autocomplete is also really pleasant for snippets and anything boilerplate heavy. The context awareness also helps. No advanced features like creating entirely new structures of files, though.
Of course, I’ll probably explore additional tools in the future, but for now LLMs are useful in my coding and also sometimes help me figure out what I should Google, because nowadays seemingly accurate search terms return trash.
Do you use any of the AI features which go for editing multiple files or doing a lot more in the same instruction?
I tried to use AI deeper, like using aider, but so far I just don't like it. I'm very sensitive to the tiny details of code and AI almost never got it right. I guess actually the main reason that I don't like AI is that I love to write code, simple as that. I don't want to automate that part of my work. I'm fine with trivial autocompletes, but I'm not fine with releasing control over the entire code.
What I would love is to automate interaction with other humans. I don't want to talk to colleagues, boss or other people. I want AI to do so and present me some short extracts.
If you are working on a greenfield project, AI codegen really shines today and there are many tools in the market for that.
With Aide, we wanted it to work for engineers who spend >= 6 months on the same project and there are deep dependencies between classes/files and the project overall.
For quick answers, I have a renewed habit of going to o1-preview or sonnet3.5 and then fact checking that with google (not been to stack overflow in a long while now)
Do give AI coding a chance, I think you will be excited to say the least for the coming future and develop habits on how to best use the tool.
Have you tried Aider?
They've done some discovery on this subject, and it's currently using tree-sitter.
We also use tree-sitter for the smartness of understanding symbols https://github.com/codestoryai/sidecar/blob/ba20fb3596c71186... and also the editor for talking to the Language Server.
What we found was that its not just about having access to these tools but to smartly perform the `go-to-definition` `go-to-reference` etc to grab the right context as and when required.
Every LLM call in between slows down the response time so there are a fair bit of heuristics which we use today to sidestep that process.
I'm using Codeium and it's pretty decent at picking up the right context automatically, usually it autocompletes within ~100kLoC project quite flawlessly. (So far I haven't been using the chat much, just autocomplete.)
I've also had some issues where it doesn't seem to work reliably, but that could be related to my setup.
not perfect but good to incrementally build things/find bugs
Founders care about development experience a lot and it shows.
Yet to try others, but already satisfied so not required.
It's amazing how much difference the prompt must make because using it is like going back to gpt3.5 yet it's the same model.
The 600 line limit might be due to the output token limit on the LLM (not sure what they are using for the code rewriting)
For the people comparing to Cursor on features, I suspect the winner is going to be hard to articulate in an A:B comparison.
There's such a difference in feel that may be rooted in a philosophy, but boils down to how much the creator's vision aligns with my own.
- arm64 build: https://github.com/codestoryai/binaries/releases/download/1....
- x86 build: https://github.com/codestoryai/binaries/releases/download/1....
> There's such a difference in feel that may be rooted in a philosophy, but boils down to how much the creator's vision aligns with my own.
Hard agree! I do think AI will find its way into our productivity tool kit in different ways. There are still so many ways we can go about doing this, A:B comparison aside I do feel the giving people to power to mold the tool to work for themselves is the right way.
So just taking these things for face value, I think there is lots to innovate. No editor (bias view of mine) has really captured the idea of a pair programmer working alongside you. Even now the most beloved feature is copilot or cursor tab with the inline completions.
So we are ways further from a saturated market or even a feature set level saturation. Until we get there, I do think forks have a way to work towards their own version of an ideal AI native editor, I do think the editors of the future will look different given the upwards trend of AI abilities.
The readme also talks about how LSP services are not exposed properly yet, my takeaway is that its not complete yet..but surely doable
If you look at the sidecar side of things: https://github.com/codestoryai/sidecar/blob/ba20fb3596c71186... these are the main APIs we use
On the editor side: https://github.com/codestoryai/ide/blob/0eb311b7e4d7d63676ad...
These are the access points we need
The binary is fairly agnostic to the environment, so there is a possibility to make it work elsewhere. Its a bit non trivial but I would be happy to brainstorm and talk more about this
https://support.mozilla.org/1/firefox/132.0.1/Darwin/en-US/x...
i'm so sorry but what do you think cursor's philosophy is
> also other niceness (like rollbacks
yep in cursor too
i know youre new so just being gentle but try to focus on some kind of killer feature (which i guess is sidecar?)
also https://x.com/codestoryai seems suspended fyi
In terms of features I do believe we are differentiated enough, the workflows we came up with are different and we are all about giving back the data (prompt+responses) back to the user.
The sidecar is not the killer feature, its one of the many thing which ticks the whole experience together.
Good callout on the codestoryai account being suspended, we are at @aide_dev
I've never understood why people say sorry for cases like these?
All our binaries are listed out here: https://github.com/codestoryai/binaries/releases/tag/1.94.2....
We also import your extensions automatically (safe guarding against the ones with Microsoft's licensed)
You can also just download in from the vscode marketplace webpage and drag and drop it in
Looks like our build pipeline is broken!
Click here to let us know?If I understood that correctly, it would mean supporting Claude via the AWS Bedrock endpoint, we will make that happen.
If the underlying LLM does not change then adding more connectors is pretty easy, I will ping the thread with updates on this.
Is it ~just about adding another file to https://github.com/codestoryai/sidecar/blob/main/llm_client/... ?
I could take a look too - another way for me to test Aide by working with it to implement this. :-)
(https://github.com/pasky/claude.vim/blob/main/plugin/claude_... is sample code with basic wrapper emulating Claude streaming API with AWS Bedrock backend.)
To test the whole flow out here are a few things you will want to do: - https://github.com/codestoryai/sidecar/blob/ba20fb3596c71186... (you need to create the LLMProperties object over here) - add support for it in the broker over here: https://github.com/codestoryai/sidecar/blob/ba20fb3596c71186... - after this you should be at the very least able to test out Cmd+K (highlight and ask it to edit a section) - In Aide, if you go to User Settings: "aide self run" you can tick this and then run your local sidecar so you are hitting the right binary (kill the binary running on 42424 port, thats the webserver binary that ships along with the editor)
If all of this sounds like a lot, you can just add the client and I can also take care of the plumbing!
My experience: 1. I didn't have a working installation window after opening it for the first time. Maybe what fixed it was downloading and opening some random javascript repo, but maybe it was rather switching to "Trusted mode" (which makes me a bit nervous but ok).
2. Once the assistant window input became active, I wrote something short like "hi", but nothing happenned after pressing ctrl-Enter. I rageclicked around a bit, it's possible I have queued multiple requests. About 30 seconds later, suddenly I got a reply (something like "hi what do you want me to do"). That's .. not great latency. :)
3. Since I got it working, I opened the sidecar project and sent my second assistant prompt. I got back this response after few tens of seconds: "You have used up your 5 free requests. Please log in for unlimited requests." (Idk what these 5 requests were...)
I gave it one more go by creating an account. However after logging in through the browser popup, "Signing in to CodeStory..." spins for a long time, then disappears but AIDE still isn't logged in. (Even after trying again after a restart.)
One more thought is maybe you got DDos'd by HN?
Yup thats cause of the traffic and the LLM rate limits :( we are getting more TPM right now so the latency spikes should go away, I had half a mind to spin up multiple accounts to get higher TPM but oh well.... if you do end up using your own API Key, then there is no latency at all, right now the requests get pulled in a global queue so thats probably whats happening.
> 3. Since I got it working, I opened the sidecar project and sent my second assistant prompt. I got back this response after few tens of seconds: "You have used up your 5 free requests. Please log in for unlimited requests." (Idk what these 5 requests were...)
The auth flow being wonky is on us, we did fuzzy test it a bit but as with any software it slipped from the cracks. We were even wondering to skip the auth completely if you are using your own API Keys, that way there is 0 touch interaction with our llm proxy infra.
Thanks for the feedback tho, I appreciate it and we will do better
Tbh I did try out their implementation and it still feels early, one of the key difference we went for was to allow the user to move freely between chat and editing mode.
Yours for example doesn't allow one to insert diagnostics or allow for example for things like inserting all tabs open and all files open at once.
They also allow jumping from editing to chatting mode by simply doing command enter.
We implicitly take in all the diagnostics on the files https://github.com/codestoryai/sidecar/blob/e5408782a3bfa461...
I won't shut up for about this, I don't understand how such an useless "feature" becomes the norm in modern IDEs.
$word is already a name for a project. Stop copying it. Change your name.
IMHO the right thing would be to use another name.
We should probably pick another name then