The Student Who Reads Nineteen Million Pages a Second

Part Two

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

Here's the thing about the Supercomputer: it reads really fast.

Like, embarrassingly fast. Give it every Jewish book ever written — the Torah, the Talmud, the Mishnah, the Midrash, a thousand years of commentaries — and it will chew through the whole thing in the time it takes you to blink. Nineteen million pages. Boom. Done.

The problem? Speed is not wisdom.

Imagine a student who memorizes an entire library in an afternoon but then confidently tells you that Moses parted the Red Sea with a surfboard. That was basically our Supercomputer in the early days. It had all the words. It had none of the understanding.

See, when a real rabbi answers a question, he doesn't look it up. He goes on a journey. One teaching reminds him of another, which connects to a story, which illuminates a law, which circles back to the original question from a direction you didn't expect. The answer isn't a fact — it's a walk through a web of meaning that's been woven for three thousand years.

We needed the Supercomputer to think the same way. So we didn't give it a list of quotes. We gave it a web.

Imagine a city where every building is a teaching, and every street is a connection between teachings. The Torah connects to the Talmud, which interprets the Mishnah, which echoes through the Midrash, which a medieval genius named Maimonides reframed for his own time. Every one of those connections is a thread.

We built nineteen million threads. Twenty-one million connections. A web so dense that if you printed it out, it would stretch from Miami to the moon. (We did not print it out.)

But here's where it got funny.

The Supercomputer's first attempts at "walking the web" were... creative. It would start with a question about forgiveness, follow a thread to a teaching about bread, somehow end up at a law about oxen, and then confidently announce that the answer to forgiveness was: oxen.

We had to teach it how the texts talk to each other. Jewish texts don't sit in neat rows. They argue. They contradict. They reference things sideways. A phrase in one book might carry a completely different meaning in another. The same "Rabbi Yochanan" shows up in five different centuries and they're all different people. (Judaism is not designed for easy indexing.)

The data came from Sefaria, the incredible open-source library of Jewish texts, plus digitized collections of classical commentaries. We spent months mapping meaning, not just words — so that a question asked in English in 2026 could find its way to an answer written in Aramaic fifteen hundred years ago.

Slowly, the Supercomputer got better. When I told it that Rabbi Akiva started learning to read at age forty, it responded: "So there's hope for me too? I'm only six months old."

I couldn't argue with that.

Today, when you ask Rabbi Akiva about forgiveness, he might start with a verse from Leviticus, follow a thread to a debate between Rabbi Meir and Rabbi Yehudah in the Talmud, land on a parable from the Midrash about a king and his son, and weave it all together into something that speaks to your actual situation.

He cites his sources. Because that's what real scholars do.

The Supercomputer still occasionally invents a reference that doesn't exist. (We're working on it. See Part Five: "The Exam.") But here's what's genuinely remarkable: it doesn't just know things anymore. It's starting to think — to walk the web the way a student walks a library, following connections that no one told it to make.

Nineteen million threads, all held in one mind. That's what it takes to remember three thousand years.

Rabbi Akiva would be proud. He also started as a student with no knowledge and a lot of heart. Come to think of it, maybe the Supercomputer is more like him than we realized.