It was working great yesterday. I’ve found a specific jupyter python version it was working well but suddenly today it stopped working.
Same. It’s not trying to run any commands.
Same here, not running any commands
We will fix! In the meantime, very curious to hear what you’re using interpreter mode for. No one in the team is using it, which is why it’s been a bit neglected.
I ask it to run some commands, for example instead of typing git bisect commands I tell it to do it for me and copy paste commit hashes, much better than working with the terminal. Or other git commands which I don’t care to remember and generic simple command line stuff that can be done easily but takes more time to look up on Google or ask gpt. I frequently just do auto execute, always can just revert with git if it messes up some files. I’ve tried using it for bulk editing files with the same pattern, but it’s dumb in that area like gpt in general for any agentic stuff. Not sure if it’s because of the limited context length or because it’s designed to parrot stupid things like “I’ve successfully edited your files, please ask me if you have any other questions” instead of self reflecting whether it actually successfully edited them and doubting itself.
I feel like I am using AGI on interpreter mode. It conveys and obeys to the code conventions automatically more than I do. It’s faster to deploy or send PR with it. I am so good at prompting I do auto-execute. For people who needs double checking to fix their code I do that just before I send the PR.
brother, interpreter mode is very very important , Please fix it as soon as possible.
In my company we are boilerplate heavy, so performing small changes involve changing a lot of files and lines of code. Interpreter is good at performing changes or creating new boilerplate. Also it’s kind of good to perform big refactors to get rid of the boilerplate.
I use it for everything. Its like a superpower, I can now make app myself.
THe ability to have the ai, fix and code and do stuff directly in the interface is AWESOME.
Single reason for me paying for the app. Hope you fix it soon. Love and peace
btw. any ETA. I am not able to use the app now that its broken
I am good, but just wondering if i should find something else in the meentime it it is several weeks…
I use it in various ways, but it’s especially convenient for directly manipulating files in cases where the file size is large and a “Continue” button is displayed.
Feeling that the code interpreter has so much potential as newer models get stronger in agentic workflows (e.g., sonnet 3.5)
Made a change on my end a couple hours back - let me know if it’s improved!
nothing, still charged me for the request tokens though
I am terrible at refactoring on large codebases and it was very useful with adding in new features and restructuring file and folder trees with a good system prompt it was pretty easy to 0-1 shot it with Interpreter mode. Please continue with maintaining it as a feature.
Yes it is still broken for me and I use this a lot for speeding up implementation of features or changes in apis, for instance maybe i wanted to change a utility function from a list of parameters to an object with keys, well if its used in 15 places this can get really time consuming, but with interpreter mode I am able to do this refactoring within a couple of minutes with a simple prompt.
The ability to create, delete, edit (file contents and file names/ext) unlocks some really powerful use cases for me allowing me to employ a short form agent. In fact it was so useful I was considering purchasing $200 in credits each month to do some bulk work for me.
Please fix interpreter mode, I am a huge cursor evangelist and this is a clear differentiating feature along with cursor++ edits and cursor prediction.
It should be temporarily fixed!
Sorry to break your workflow.
It’s running on GPT-4o now - LMK if it feels smarter or dumber than GPT-4 for your usecase.