Even in the music case, a much more helpful framing for me—that I learned way later than “relax and have fun”—was “your brain isn’t good at telling excitement for a show from being chased by a tiger,” since that at least acknowledged the fight or flight feelings.

This has only worked for me in the case of stage fright when playing music in front of a crowd, and even then it was more reminding me that music was something that I already had fun with. In other situations I haven’t found it help much.

Absolutely, and it’s the kind of thing I find myself coaching people on the senior->staff track about. It’s rarely “go learn this tech thing,” it’s “go learn how to tell a story that someone wants to hear” or “learn some basic economics around incentives“.

Would still listen, and I don’t think they’re necessarily mutually exclusive.

Working like this also makes it way easier to “sell” your projects to leadership. The people you’re helping will go to bat for you, and VPs love when you can name their reports whose problems you’re solving with a project.

An ancillary benefit is that you’ll also become known as a trustworthy person to bring issues to, which is very useful when you’re launching the updated version of that new not-so-simple-anymore tool.

And then you do it again. And again. And again. Eventually people will have similar-but-different problems and you’ll have to expand a tool‘s capabilities. Do that enough times and you’ll have a good shape for a tool that solves a class of problems in a way folks actually need.

The right way is to ask a single engineer what problems they have, and find the simplest solve for one that’s high leverage. Maybe it’s a script. Maybe it’s patching some existing tool that almost does the right thing. Maybe it’s just showing them how to use a tool and updating your docs.

The wrong way (which I’ve also done plenty of) is to ask a bunch of people about their problems, generalize them until they’re all a similar shape, and come up with a complex system that solves that general class of problem that everyone has in the abstract but no one has concretely.

And I mean that very literally. In my experience—I’m currently the most senior engineer in a developer experience org for a company of about 500 engineers, for whatever that’s worth—the best way to make great tools involves asking that question a lot.

Fun concert or very unfortunate goldfish.

Yeah, though mostly split between fiction and non. The rare time it’s been more than two at once is I read a novella for a few days straight or something.

Now we just need a Claude Code plugin that’ll play Unchained Melody

A teammate of mine did that program and really enjoyed it. Congrats and best of luck!

When I last tried Claude Desktop a few months ago I liked the interface, but had a lot of issues with the sandbox not being the same as normal CC. Have you found that as well? Is it just a process of setting it up as you go?