Every founder I know can tell you what they built last quarter. Almost none can tell you what they refused.
That asymmetry is the tell.
Strategy, properly understood, is not a plan for action. It is structured sacrifice — a ledger of forgone options, kept as carefully as any P&L.
For most of computing history, restraint was enforced by scarcity. Compute was expensive. Engineers were expensive. Friction did the choosing for you.
That friction is gone. An agent can scaffold a product overnight for the price of a coffee. The marginal cost of building has collapsed toward zero, and with it, the old excuse for not building has collapsed too.
Which means the question that used to be optional is now existential: what did you choose not to do, and why?
Not the roadmap. The graveyard beneath the roadmap.
Michelangelo's line about the statue already existing inside the marble is usually quoted to celebrate vision. It is really a description of subtraction — everything he declined to leave in the stone. The sculpture is the residue of ten thousand refusals.
Founders today are handed infinite marble and infinite chisels, all algorithmic, all nearly free. The temptation is to keep carving because carving feels like progress. Motion is not direction, and an agent will happily generate motion all night.
Here is the inversion nobody wants to sit with: doing has become the cheap, safe, almost cowardly option. Not doing — the deliberate, defended, un-glamorous act of leaving a feature unbuilt, a market unentered, a demo unshipped — now carries the higher risk, and therefore the higher signal.
When everyone can build everything, the scarce resource is no longer capability. It is judgment about which capability to withhold.
Think of it as a second ledger, kept beside the visible one. Every shipped feature has a shadow entry: the three things not shipped so this one could be. Nobody audits that ledger. Everyone should.
The cost of tokens keeps falling. The cost of your attention does not — it is the last commodity on earth with a genuinely fixed supply, and every AI agent is, structurally, a bid for a piece of it.
So the winning position, paradoxically, starts to look like stillness. Not paralysis — a chosen, defended absence, the kind a sniper holds before a shot, a market-maker holds before a trade. Stillness that is a position, not the lack of one.
Ask a founder what they built and you get a résumé. Ask what they didn't build and you get a worldview.
The cursor blinks on an empty terminal. Nothing is running. That, increasingly, is the most expensive-looking thing in the room.