The Oracle’s answer

Lecture notes

This is a reading preview for the opening lecture. The recording and its transcript will be added after production.

Before Croesus of Lydia went to war, he did what a careful leader should. He tested his instruments. In Herodotus’s account, the king sent messengers to the great oracles with a single question: what was he doing on a particular day? Delphi answered correctly. Then he asked the question that mattered. If he made war on Persia, the oracle replied, he would destroy a great empire.

He attacked. A great empire fell. It was his own.

The story is often read as a warning about ambiguous prophecy. I read it as a warning about something narrower and more useful: the difference between an answer and a decision.

The oracle's answer was, in its way, correct. Croesus had even validated the instrument; he ran what we would now call an evaluation, and Delphi passed. His error was not trusting a bad source. It was letting a good source do the one job it could not do, which was to decide what the answer meant for him.

Every leader now keeps a Delphi on the desk. It produces fluent answers, often quickly enough to hide the interval in which a person ought to think. It may pass our evaluations. And it recreates the old danger in a new form: answers so confident and well phrased that they feel like decisions.

They are not. A model can tell you that a market will contract, that a contract hides a risk, that a candidate fits a profile. It cannot tell you which empire you are about to destroy. That requires knowing what you are protecting, what you are willing to lose, and who will bear the cost if you are wrong. Those are not facts in the world. They are commitments you hold.

Herodotus gives the story a coda. When the defeated king sent to Delphi to complain, the reply was blunt: he should have asked which empire was meant. It is the most modern sentence in the book. The failure was a missing question, the one that would have tied a general answer to a particular life.

The practice this week

Before you act on a model's answer, ask the three questions it cannot answer for you:

1. Which empire? Name exactly what this decision could destroy (an asset, a relationship, a promise) and whose it is.
2. Who answers for it? Name the person who will stand in front of the consequences. If it is no one, the decision is not ready.
3. What would change my mind? Write down, before you see the outcome, the evidence that would reverse you.

None requires a new technology. Each forces a distinction Croesus failed to make: between receiving an answer and choosing an action.

From the Canon

Herodotus, Histories, Book I, the Croesus chapters, in G. C. Macaulay’s translation.

Nikos