would it not be usable to have completion be case insensitive? I seem to be able to use that… if I only remember “something with down”, I could just as easily forget the capitalization of “down”. maybe I have downloads and Down? why not show everything matching case insensitively and let the user decide what’s the correct one?

I didn’t really understand what you thought the regex did incorrectly, but I think the regex works fine, at least for most implementations, anyways what I meant is just a case insensitive version of the regular substring completion, which shouldn’t be too difficult to make.

The only thing it solves is the frustration of having to look for a file/directory twice because you didn’t remember it’s capitalization. again, those are different characters just like a do and downloads are different strings, but it can be easier for users if they can just press tab and let the computer fill the part of the name the don’t remember (or don’t want to type).

you don’t need an advanced algorithm or and AI, there are many easy ways to make completion case insensitive (like that regex for example). Issues involving names are inherently somewhat linguistic, but either way interactive shells are meant to be (at least somewhat) usable to humans, and as seen by the post, some people would prefer completion to be case insensitive.

but why do we have to match specifically against substr*? it’s not a law of nature, we could also match against the regex (?i)substr(?-i).*

not saying that one option is necessarily better, but I don’t see a good reason for which any one of these options would be terrible

You could also say that down should not complete to download since those are completely different strings and you shouldn’t expect one to get you the other.

There is truth in this, but it isn’t as true as some people seem to think. it’s true that trial and error is a real part of working in ml, but it isn’t just luck whether something works or not. We do know why some models work better than others for many tasks, there are some cases in which some mannual hyperparameter tuning is good, there was a lot of progress in the last 50 years, and so on.