Something that’s helped me a lot in figuring out AI: Building a sense for when Claude is ready to go off and do its thing vs needs more conversation with me.

It almost feels like a new soft skill for knowledge workers in the ✨ age of AI ✨

Sparkle emojis aside, it might’ve become the most useful thing I’ve noticed myself doing to unlock some more speed (and take on things I couldn’t as easily before) without burning myself or other people out.

Early on I struggled with this seemingly now-classic problem: I’d have one convo going on in Claude Code, then switch to another, realized I need to jump back to the first, open a 3rd in Claude Cowork to tackle something while waiting to hear back on the other two, and so on, and that was just pure chaos.

Over time I started to notice the kinds of things Claude would ask more questions about, which in turn helped me notice more finely when I left uncertainty or openness to workshop something my prompts to it. This helped me realize when I should stay focused on sharpening the task I wanted to send Claude off to do, or when I felt comfortable just sending it off.

I now feel a bit closer to a centaur (with extra appendages that enhance what I can do), rather than a reverse-centaur (at mercy of the machine with all the appendages that I’m scrambling to keep up with).

This isn’t just about being a 🧠 master prompt engineer 🧠 in fact, some of the most fruitful outputs I’ve gotten from AI tools come from deep, substantive conversation about a problem space I’m working through.

But rather from actively resisting the temptation to switch contexts too much, giving each session the focused attention it deserves, and knowing (with experience over time) when Claude is ready to work. And then you can move onto the next thing, or better yet, go touch some grass 🌱

Again, it feels almost comically simple, but that’s what the major unlock has been for me. It’s an important one in light of the seemingly popular feeling that AI is making work more, not less intense.