[diziet]

I would also love to see a token budget use for the chats -- to know when the model is about to run out of context. It's crazy this is not there.

[joquarky]

It would have to be intentionally vague since there is no hard cutoff threshold.

[bspammer]

Claude code does it, to a precision of 1 digit of a percentage. That’s more than enough to be useful.