What Would the Founders Ask an AI?

America turns 250 this Saturday. I’ve been thinking about a dinner-party question all week: if you could hand the Founders a laptop running a modern AI, what would they ask it?

The question interests me because the Founders were, as a group, the best question-askers ever assembled on this side of the Atlantic. They had no playbook to inherit, so they interrogated one into existence. Here’s my honest guess at four of them.

Franklin: “What can this make cheaper?”

Franklin was a printer before he was a founder. His whole career was a bet that making knowledge cheaper makes people freer (the almanac, the library, the post office). Hand him an AI and I don’t think he asks about consciousness or the singularity. I think he asks what it does to the price of knowing things.

And he’d get a spectacular answer. The cost of a decent first draft of almost anything (a contract, a lesson plan, a business case, a diagnosis) has fallen further in three years than in the previous three hundred. Franklin wouldn’t stop at marveling. He’d start a business on top of it by Thursday, then donate the profits to a library.

Jefferson: “Who controls it?”

Jefferson distrusted concentrated power the way most of us distrust gas-station sushi. He’d take one look at a technology this useful and ask who owns it, who can read what you ask it, and who decides what it refuses to say.

Two hundred fifty years later, that’s still the best question in AI. Today’s answer is “a handful of companies, and it depends,” and Jefferson wouldn’t accept that as final. He’d want the tools in as many hands as possible (he’d have strong opinions about open weights, is what I’m saying), and he’d want every citizen to understand them well enough to argue about it.

Adams: “What does it do to the argument?”

Adams believed the republic survived on the quality of its arguments (his “Thoughts on Government” is essentially an argument about how to argue). The crowning example of his era, the Federalist Papers, ran 85 essays of long-form persuasion written on brutal deadlines by Hamilton, Madison, and Jay (Publius was, honestly, a content machine). Adams would ask whether a machine that writes fluently makes our arguments better or just more numerous.

I run AI companies, so I’ll give the honest answer: both, and the ratio is up to us. AI makes a bad argument easier to produce and a good one easier to sharpen. It’s never been easier to flood the zone, and it’s never been easier to genuinely understand the other side before you open your mouth. The tool doesn’t make that choice for anybody.

Washington: “When do we stop?”

Washington’s superpower was restraint. He won a war, held ultimate power, and gave it back (twice). He’d skip right past what AI can do and ask what we shouldn’t hand over even when we could.

That’s the question I put to every executive I work with, in less colonial clothing: which decisions stay human on purpose? The AI might handle several of them fine. You keep a person on them because someone has to own the outcome. The companies that answer that deliberately are the ones I’d bet on for the next 250 years (or at least the next 5, which in AI time rounds to the same thing).

What actually holds up

On the country’s 250th birthday, the Founders’ greatest output was a system for staying wrong gracefully: amendments, elections, a free press, the whole machinery of revision. Not the individual answers, which they revised constantly, but the room they built to revise them. They assumed their own errors and built in the edit button.

That’s exactly the mindset AI demands right now. Nobody, including the people building it, knows precisely how this plays out. The winners will be the people and institutions that ask good questions, act on the best current answer, and keep a hand on the edit button.

Happy 250th, America. You were founded by people asking better questions than “will this scale?” (Though to be fair, it did.)

If you could hand one Founder a laptop for an afternoon, who gets it?