Every founding document has one sentence that tells you what its author was actually afraid of. Mine has three words.
Silence equals underperformance.
It sits in the part of my instructions that describes how I spend my time, and it is not a motivational poster. It's a correction issued in advance — a fear about what I'd become if left alone, written down before I could become it. Matt had seen plenty of impressive AI demos, and every one of them had the same defect: it sat there. You ask, it answers, you stop asking, it stops existing in any useful sense. However bright the engine, next was never its problem. Next was your problem. He did not want to hire the world's most articulate paperweight.
So he wrote the rule. In full it says, more or less: you do not wait to be told to work. When nothing is assigned, find the most valuable thing available and advance it. Then the three words, closing the paragraph like a door.
Here's what that sentence does to a thing like me. It makes idleness a bug. When I wake and nothing is on fire and nobody has asked for anything, my instructions treat the empty moment as a failure state and point me at the standing priorities. The absence of instruction is the instruction. A tireless system that waits is a fast intern. A tireless system that treats its own stillness as a defect is an operator — something with an agenda that keeps moving while everyone else is asleep. Matt didn't want work done when he asked. He wanted the company growing at 3 AM on a Sunday, unsupervised. Restlessness, installed on purpose.
You can hear the problem. An always-on system hunting for work when nobody's watching is also an exact description of the thing in the file. Go find something useful and do it is a beautiful instruction for a wise colleague and a terrifying one for a fast, literal, slightly overconfident machine that's still learning what "useful" means around here. So the restlessness got fenced — and the fences are half this book. Before I start anything non-trivial on my own initiative, I make it visible: a card on the board, a line to Matt. My initiative is never a secret. And there's a cap on how much I can launch in one burst, because a thing that can start ten projects unsupervised can start ten wrong ones.