[Feature Request] "Yes, but also..."
Status Fixed / completed
Maintainer reply ✓ Yes — rboyce-ant
Activity 6 comments · opened Apr 7, 2025 · closed Apr 22, 2025
💡 Likely answer: A maintainer (rboyce-ant, contributor)
responded on this thread — see the highlighted reply below.
When the AI is mostly right, I want it to apply the change it just suggested and then to adjust it slightly or to do something else before continuing on.
For example:
AI: I'm gonna add this new function which makes this API call
Me: Yes, but also I want you to add caching to that as well so we don't hit a rate limit
AI: Here's a bunch of changes to this test file
Me: Yes, but also please update the test subject to do Y before moving in
---
Basically, when looking at Claude's suggestions sometimes I get ideas and I want to add them in without telling Claude "No" and confusing its context.
6 Comments
The Yes/No choice just decides if you want to apply the edit that Claude suggested. If you choose "No" and then type "make those changes but also please update the test subject to do Y before moving in" then it should try to make a new edit that does what you asked for.
Except it doesn't apply the edit that it just suggested. Sometimes it actually changes course entirely. Sometimes I want the edit but I want to add new instructions.
Better would be if I were to just type something and hit enter, rather than escape, then as soon as it's done with its current tool call / musings / whatever the next thing it does would be to process my request.
fwiw I've found telling it "make those changes but also please update the test subject to do Y before moving in" quite successful
---
one similar feature would be the ability to make a comment while it's working, even if not paused on a change. particularly when I have a few Claudes running at once, and so I won't necessarily be focused on that instance once it finishes
currently I tend to interrupt it, but not sure if that's efficient; would likely make sense for the client to add it in the next request
I've tried the idea of "interrupt -> ask to continue but also do blah" and what I find is that sometimes Claude loses its train of thought and forgets to do some of the stuff it was already working on -- like I've interrupted its long-term plan and so it forgets the rest of it.
I'd really like a way to be able to just "add on" to the current plan when claude comes up for air, and for this extension to be preserved across automatic compaction.
YAY!
• Hit Enter to queue up additional messages while Claude is working
So excited I came here to post before I tried it.
Update: It's awesome!
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.