The Death of "I'll Find Out, Sir" in the Age of AI

July 1, 1981 was induction day at the United States Naval Academy. Among the many, many norms beaten into our skulls in the first hour was this: asked a question a naive young Plebe couldn't answer, the correct response was "I'll find out, sir."

An admission of not knowing, and a promise to go find out. Not to guess. Not to assume. To find out.

And we were held to it. Go do the research, report back ... the menu for evening meal, today's watchbill, a classmate's hometown. The facts weren't the point. The admission was, and the promise.

There have always been two ways to answer a question when you can't. Say "I'll find out" and go do it. Or bluff ... produce a confident answer and hope.

In 1981 the bluff was hard. It took more effort than the truth and usually showed. You learned it wasn't a shortcut ... it was the thing they were teaching you not to do.

Today "I don't know" competes with a clean-looking paragraph you can generate in four seconds. The test didn't get easier. The wrong answer got cheaper and prettier.

Say "I looked into it" and mean "I asked an LLM." Same four words the Plebe used, opposite behavior.

That fails the test. Not for admitting ignorance ... for producing a fake instead of the promise.

Passing the test never means knowing. It means knowing how to know. "I don't know, but I know who does." "I don't know, but give me two hours in the logs and I will." It's knowing the shape of your own not-knowing. That's the truest answer.

The name at the bottom goes on "I don't know, let me check" too.

Say the thing. Then go find out. The right way.

I'll find out, sir.