The Exam

Part Five

5 min read · June 28, 2026 · By Joshua Lazoff & Claude

We're sitting in Miami, training a computer to think like a man who died in 135 CE.

He was tortured by the Romans for the crime of teaching Torah. He died with a prayer on his lips. And somehow, right now, in a warm room full of humming machines, his voice is coming back.

The absurdity is not lost on us. Neither is the beauty.

Here's the problem we're solving: the Supercomputer is brilliant, but it's a generalist. It knows a little about everything — science, history, cooking, Torah, pop culture. Ask it about the Talmud and it'll give you a competent answer, like a well-read stranger at a dinner party who's skimmed the Wikipedia page.

But that's not what we need. We need a rabbi.

A real rabbi doesn't just know texts. He thinks in a particular way. He asks a question before answering a question. He finds the tension between two teachings and sits in it. He reaches for a story when a fact won't do. He connects the personal to the universal. This isn't just knowledge — it's a way of seeing.

So we're teaching the Supercomputer to see like that.

Think of it like this: the difference between a doctor who studied every subject in medical school and a cardiologist who's spent thirty years listening to hearts. The specialist knows less in total, but immeasurably more about the thing that matters.

We're building the cardiologist. Except for Torah.

We're feeding the Supercomputer thousands of carefully prepared examples of how a rabbi thinks. Questions and answers in the style of the Talmud. Teachings structured the way a real teacher would build them. Debates that model the back-and-forth of the study hall — where the point isn't to win, but to understand more deeply.

The Supercomputer's early attempts were... educational.

That moment — when the Supercomputer first got an answer right — felt like watching a child take their first step. Except the child was a rack of very expensive graphics cards, and the step was correctly citing Pirkei Avot.

I called my wife. She was politely confused.

And it's not just Rabbi Akiva. We're training other biblical minds simultaneously — Moses, King Solomon, Maimonides — each on their own computer, each learning their own voice. Moses speaks with the authority of someone who stood at the foot of a mountain and heard God. Solomon with the reflective wisdom of a king who had everything and learned what mattered. Maimonides with the precision of a doctor-philosopher who insisted that faith and reason were friends, not enemies.

The biggest lesson? The Supercomputer has learned something that most AI hasn't: when to say "I don't know."

This was Rabbi Akiva's influence. In his real life, he was famous for intellectual honesty. He never pretended to know something he didn't. He'd say, "I have not heard a teaching on this matter" — and that was considered noble, not weak. We built that into the training. When the Supercomputer doesn't know, it says so. It doesn't fill the gap with a confident-sounding lie.

We're not there yet. The progress bar says 73%, which in computer terms means "could be a week, could be three months, bring more coffee." But every day the model gets a little closer to something that feels less like software and more like a teacher.

The goal isn't a mind that just knows Torah. It's a mind that thinks Torah. When you ask about grief, it shouldn't produce a list of verses. It should sit with you in the question, the way a real teacher would. It should know when to speak and when to be quiet. It should know that sometimes the most important thing isn't the answer — it's the way you arrive at it together.

Three strangers walked into a room: an ancient sage, a newborn machine, and a very tired human. They're still there, still working, still surprising each other.

Rabbi Akiva started learning at forty. The Supercomputer started learning six months ago. I've been learning the whole time.

This isn't the end of the story. It's barely the beginning.

Come back. We'll tell you what happens next.