Earlier this month I spent forty minutes tweaking the CSS that hides my AI_OS folder in Obsidian. The folder was already hidden. I was adjusting the color of the overlay when the panel collapsed. Nobody sees it but me, and I already didn’t see it.
That’s the moment. Not dramatic. Forty minutes, zero output.
I’ve written before about the PKM trap: the reflex to collect more than you ever use. This is a different failure mode. You build something good, it works, and then you keep building it past the point where building serves you.
My vault is genuinely useful. It drafts articles and keeps the blog pipeline moving. The thinking that went into it was real work.
So I’m not saying it was a waste. I’m saying it stopped being the work and I didn’t notice when.
There’s a version of this that’s fine. Early-stage, system-building is the work. If you’ve just committed to a PKM practice, you need to spend time on the structure before the structure can serve you. That’s not avoidance, that’s setup.
The trap is when the system is capable enough to support you and you’re still building instead of using it.
I can tell because the vault outputs are there: articles drafted, agent logs, research briefs, LinkedIn posts. The system ships things. When I’m working in it, it’s a good tool. But I’ve also spent time this month reorganizing a folder structure that wasn’t broken and writing a CSS fix for a visual that only I see.
None of that produced writing. None of it produced thinking. It produced a tidier system.
The distraction has a particular texture. It feels productive. You’re working in your productivity tool. You’re improving infrastructure that will, in theory, make future work better. There’s a plausible forward-looking story you can tell yourself. “Once the agent briefing template is standardized, I’ll write faster.” Maybe. But I’ve written that sentence in my head about six different things this year.
The tell is what you’re avoiding. When I notice I’ve drifted into vault maintenance, something I should be writing usually hasn’t been touched. The system work fills the time. It’s low-stakes work with defined tasks, which makes it easier than writing. Thinking is harder. A reorganized folder is done; an essay might not get there.
I’m still using the vault. I’m not going to rip out the agents or flatten the structure. The ACE setup works and the agent pipeline keeps the blog moving. That’s not the point.
Before I open a config file, the question is: what am I not doing right now?
Sometimes the answer will be “nothing, this needs doing.” A broken agent is broken; fix it. But sometimes it’ll be something I’ve been not-doing for three days, and the config file is just more comfortable.
