I have a tinkering list. An Obsidian database full of ideas I could build with Claude Code. Tools, automations, small prototypes. I used to write things like this down too, but for a different reason. Back then it was to not forget them — because building anything took so much time that most ideas just stayed on the list. Now I write them down to stop myself from doing them immediately. I keep hearing it from people around me: they've gotten a bit addicted to Claude Code. Not in a worrying way, but in the sense of: you get things done so fast that you want to keep going, and going, and going again. Then I watched the video "Why Everyone is OBSESSED With Claude Code." The creator analyzes Claude Code through game design principles. Short feedback loops, visible progress, unpredictable rewards. "Just one more prompt" feels exactly like "just one more turn." I recognize that. Jarno Duursma wrote about it too — his work feels more intense. "This really can't go on like this," he says. But it's not just those addiction mechanics. There used to be friction in execution. You'd only start something if you thought: this is worth my time. Now that threshold is gone, and every thought wants to be built out — because you don't have to do it yourself anymore. It's a new kind of **ITCH: Ideas That Crave Happening.** An idea that used to disappear because building it cost too much now keeps itching. Because you know you can scratch it. You know that with an hour or so, a pretty decent prototype is within reach. We work faster, but the filter is gone. We wanted the friction removed. Now that it's gone, we notice what it did. The brake was the filter. Does that mean we now get better-developed bad ideas? My solution comes down to two things. **1) The tinkering list:** a place where I park ideas instead of scratching them immediately. When I have tinkering time, I pick the most interesting one — or the one that still seems worth it after sitting with it for a while. The rest stays parked. **2) AI-free time.** I thought I could manage it — give Claude a quick task, then get back to my real work. But I know that doesn't actually work. Attention residue means I keep drifting back to the Claude task anyway. I think: just one more prompt, then it can keep running. But then I'm left with fragmented time for everything else. So I set aside blocks: time *with* AI, and time *without*. At the PKM Summit panel, four of us were asked what stood out to us. There was a lot of talk about AI, but I talked about time. At that point I thought — hmm, am I saying the right thing here? Talking about AI tends to land much better with audiences. But no, I talked about time. Time is powerful. And building in waiting time is going to matter more and more — waiting time to find out which idea is still floating after a week. So: I have a tinkering list. Do you?