The tool that started working against me
Every founder eventually builds something useful enough that it begins generating its own busywork.
I pulled up a follow-up tool yesterday to check something routine. Seventy-seven open items. That sounds like a real backlog.
Then I looked at where they came from. Sixty of the seventy-seven were messages the tool had sent to itself. One original question had spawned twenty-five rows of follow-up. The system had been diligently chasing its own outputs for weeks while the actual open work sat waiting.
The machine was organized. The machine was busy. The machine was producing nothing.
Infrastructure that feeds itself
This is a failure mode I don't see named enough. You build something to extend yourself - to flag what hasn't moved, to follow up when you forget, to make sure nothing falls through. It works. So it keeps working. Past the point of usefulness, it starts treating its own activity as new signal. Chasing the chase. Following up on the follow-up.
I've watched founders do this with their whole businesses, not just their tools. The weekly review that generates more to review. The productivity system whose maintenance is its own biggest task. The board update that triggers six clarifying emails. At some point the infrastructure stops serving the work and starts competing with it.
The fix was simple: close the loops that should have been closed weeks ago, tighten what counts as open, add one rule that stops the system from treating its own outputs as new inputs. Twenty minutes. Done. Lean again.
But the lesson I'm sitting with isn't about the fix. It's about how long this ran without a signal. Nothing looked wrong. The numbers looked fine. It was only when I asked where each item came from that the ratio became obvious.
Steve Jobs said real artists ship. True - but shipping is a decision, not just a motion. It means knowing whether you're moving real work or running in place while the dashboard shows activity.
The question worth asking at the end of any day: if I stopped checking this thing tomorrow, what real work would fail to happen? If the answer is nothing, I've been maintaining the maintenance.