Two thousand
years.

Of trying.

By 1970, it had been done.

Done what?

Gold. From lead.

Try it.

An interactive story by Worth Building.

0 years gone

Perhaps a little more heat.

Ask a different question

The lead is still lead.

Each turn stands for forty imagined years. The lives and attempts are fictional, not a historical timeline. The search itself lasted centuries.

What happens to air
when you squeeze it?

Try doubling the pressure.

The same air. Less room.

12.00 mL of space.

Return to the same pressure. You’ll get the same result.

A model of Boyle’s law, not his apparatus: a fixed amount of ideal gas at constant temperature. These are calculated values, not experimental measurements.

Both questions occupied
the same people.

Robert Boyle studied gases with an air pump and helped establish experimental chemistry. He also thought metals might be transmuted into gold. The experimentalist and the alchemist were the same person.

Alchemy and chemistry overlapped for centuries. There wasn’t a morning when everyone put away the symbols and agreed to become scientists. Even the words were once used interchangeably.

The useful distinction is something you can ask of a claim today.

“Trust me.”A result you have to take someone’s word for.

“Try it.”A method someone else can check.

“Nearly there.”No way to tell whether the next attempt is any better.

“Here’s what changed.”A record that includes the attempts that didn’t work.

Boyle’s experiments and beliefs, at the Science History Institute.

About the gold.

In 2025, the ALICE collaboration reported measurements of lead turning into gold at the Large Hadron Collider. Removing three protons from a lead nucleus leaves a gold nucleus. Chemical reactions cannot do that. Nuclear physics can.

The gold nuclei hit the accelerator’s beam pipe or collimators and broke apart within a tiny fraction of a second. Across the LHC’s second run, the amount produced at its four major experiments was about 29 trillionths of a gram.

The jeweller can probably take the afternoon off.

How ALICE measured it.

The earlier experiment used a lead target. Its mercury products decayed into radioactive gold. Hansen and colleagues, 1970.

“My best guess is divine benevolence. […] It’s more like alchemy or whatever chemistry was in the Middle Ages.”

Noam Shazeer, on AI.
Fortune, August 2, 2024.

There’s work to do, then.

There are things neural networks do that researchers are still trying to explain. That leaves room for experiments small enough to follow all the way through.

For example, researchers studying small models in 2022 examined how a network can represent more features than it has dimensions. Their article includes interactive diagrams and code. A small model doesn’t explain every large one, but you can inspect the result and try changing it.

Pick a question you can test. Write down what you expect. Change one thing. Keep the result, especially if it surprises you.