Seneca, the wealthiest private man in Nero’s Rome, kept a standing instruction to his own household. Set aside a few days each month, he wrote to Lucilius, and live as if the wealth were already gone. Coarse food, a hard bed, the plainest cloak. Then ask yourself a single question. Is this the condition I feared?

He was not preparing to lose his fortune. He was preparing to not be ruled by the loss if it came. The Stoics had a name for the practice. Premeditatio malorum. The deliberate rehearsal of the bad thing, in advance, in full.

He had no cloud vendor. He had the same problem you do.

In May 2025, a founder in Jakarta woke to a sign-in screen that no longer worked. Her company verified its users through World ID, the identity layer Sam Altman’s Tools for Humanity had pushed across Southeast Asia. On the fourth, Indonesia’s Ministry of Communication and Digital suspended Worldcoin and World ID, citing a registration the operator did not hold. Overnight, by memo, her single point of entry was gone. Not degraded. Gone.

The dependency you cannot name is the one you have not rehearsed losing.

She had built a real company on one input a government could delete before breakfast.

This is the position more founders are in than admit it. The dependency is rarely the thing on the pitch deck. It is the quiet one. The model API from a single foreign lab. The payment rail that clears through one corridor. The cloud region that sits inside one jurisdiction’s reach. In January 2025, Italy’s data regulator pulled DeepSeek from the app stores in a morning. Every product that had wired itself to that one model found out, that day, what it had actually been standing on.

The pre-mortem is the discipline that names the thing before the regulator does.

It is not the same exercise as rehearsing a bad quarter. A bad quarter is slow and arrives with warning. The vanished vendor arrives with none. The pre-mortem for it is colder and more specific. You sit down, before anything is wrong, and you ask the question Seneca asked of his own wealth. If this one input were switched off by morning, what breaks, in what order, and how many hours do I have.

You write the answer down. The vendor that has no second source. The customer whose contract assumes the vendor. The week you would lose finding a replacement you should have found already. The founders who ran this exercise in the quiet months were not spared the ban. They were spared the panic. They had the migration half-built and the customer email half-drafted before the memo existed.

He robs present ills of their power who has perceived their coming beforehand.

The ones who had not done it spent the first day discovering, in real time, what they were made of. That discovery is the most expensive way to learn it.

There is a hard version of this and a comfortable version. The comfortable version is a risk register, reviewed quarterly, that nobody reads. The hard version is naming the single dependency you most do not want to lose, the one you have quietly decided is too central to fail, and rehearsing its loss anyway. That is the one. The dependency you cannot bring yourself to imagine losing is the one you have not rehearsed losing.

Seneca’s whole point was that the rehearsal removes the disaster’s power to surprise you. The imagining is a tax you pay early so the event, when it lands, costs you only money. The founder in Jakarta who had war-gamed the suspension slept through the week it happened. The one who had not did not sleep at all.

He robs present ills of their power, Seneca wrote, who has perceived their coming beforehand.