One weird trick to making Claude Code palatable to use

I'm in awe of Claude Code, but I also curse it many times a day. One thing that I've recently started doing that has made a big difference to my mental health and made me enjoy using it more is this dumb trick... simply ignore what it says at the end of a task.

The self-congratulatory backslapping or glazing of the programmer, e.g. "Perfect! The thingamajig now frobbles before it frazzles, making this a much better experience for the user and improving the versatility of the system" (and of course accompanied by several lines with green checkmarks)...this drives me nuts. Why? Because it told me it's done all these things, so I adjust my mental model of the code accordingly. But then it turns out it actually hasn't done all those things at all.

If this was a junior dev you'd given a task to, and they came back full of praise for themselves for the stellar job they'd done - and then it turned out they'd botched it badly, after a few times you'd be having an HR discussion.

So instead, let all that stuff at the end of the task scroll by, train your eyes to ignore it, and just test whether it's working as intended. Claude Code is no longer my buddy, it's now a tool. Sanity is restored.