Claude Code used to stream the thinking process in verbose mode. Now that has been replaced with "transcript mode" which doesn't actually give much more information and also doesn't stream anything. They also recently removed (in certain situations) the counter of how many tokens the model's generated in its response in progress, so the only way to tell if it's stuck is to wait 10 minutes and then retry.

Sure, I can read the diffs as they're generated (and I do). But proper transparency goes further than that, and it's being stripped away.