Coding agent in the IDE: Generate task plan and get plan-aware code completion
twitter.comHey HN, CodiumAI co-founder here.
We’ve come up with a bit of a different concept for what a coding agent should be. We believe it should work in tandem with a developer inside the IDE. Over time as the tech improves, it will get more and more autonomy.
We’ve been using our coding agent internally and see a 5-10x boost on some tasks.
The agent is available now to Codiumate VS Code users. We want to hear what kind of tasks it works well on and improve it over time to expand the task set. Would love to get feedback.
https://marketplace.visualstudio.com/items?itemName=Codium.c...
> internally and see a 5-10x boost on some tasks.
That's my experience with co-pilots too:
- Generating tests
- Generating functions consistent with prevailing style of similar functionality in the existing codebase. The greater the consistency, the more helpful the AI is at generating.
- Telling me why my code is crap by adding a `# todo: ` above some code and seeing what the AI suggests should be changed :)
What other tasks do you see as good target for 5-10x boosts?
How is it different than other agents and co-pilots?
Well,
To put it simple: Like copilot, it does have an auto-complete and also chat interface. Different than copilot, it focuses on generating a full code task plan, then having the auto completion work acrroding to your plan, and it check the of quality code.
Like agents, it tries to help you complete a full task, yet,does that in tandem with you, as you work inside tour favorite IDE writing the code with you. In addition, there is focus on code quality, testing, and fetching relevant context from your codebase.
As long as your favorite IDE is VSCode?
We will support other, I promise
it's a good start. VSCode is most popular. Largest impact. More potential users. More likely to find a customer base.
I like this. I think AI agents are the future of software development.
AI will not replace devs. Devs that use AI will replace devs that do not use AI.
The most effective devs will be those employing a fleet of AI agents, acting as the glue and guiding hand for what the agents should produce.
This helps us get to that future, so I think this has legs.
I use VS code. I will try this out.
I tried Codeium out last week and couldn't get on with it at all. I was using their Sublime Text integration which I have to say is barely integrated - it was slower to suggest code than CoPilot, the suggestions weren't as useful, the way it just put the suggestion directly into the buffer you're typing in was extremely distracting. I turned it off after about 5 minutes. I suspect the Sublime integration has had the bare minimum of work needed to be able to check the "works in Sublime" box on the marketing site, and there's no way I'm going to use VS Code or a fork of it instead of the editor I live and work in.
Oh that's confusing, I'm writing about Codeium, and this post is about Codium.
What else will it use as a context? Can it be controlled by the user or is done automatically like other auto complete extensions?
The user can choose the relevant context as follows: go to the relevant code part/snippet/files, select, then Ctrl+shift+E, one by one.
In the near future, the agent will suggest these for you, after first indexing your code base
Which model/s are you using?
SaaS uses best of bread from OpenAI, whole enterprise get to choose also from our own model for on prem deployment