Can we get a little more data about how /edit is different than command+k?
does the model that it’s using is powerful as gpt4? (since you guys said it’s faster)
We expect /edit to be as powerful as cmd-k with gpt-4. If you notice any discrepancies, please let us know. Note that it respects your model selection (i.e. it uses gpt-3.5 if you have gpt-3.5 selected).
Cmd-k works by asking the model to regenerate the entire selection. The /edit feature only asks the model to generate what should be changed, which is why we expect it to be faster. The hard part is figuring out where to edit. This is where we’re using non-public models — not sharing specific details on this yet, but there may be a blog post on it at some point!
Sorry if I’m being dense, but I’ve never heard of “/edit”. Where can I learn even just the basics of what it is? Is it a nightly build only thing?
No worries — it is nightly-only for now! Coming to stable in the next build (should be this weekend).
The idea is that you can type “/edit” in the chat to edit an entire file at once based on your chat history.