The Miracle That Wasn't

That was it. That was the miracle. A rock leaned.

For about three weeks in the summer of 2023, that tilting rock was the most important object on the planet. On July 22, researchers at the Quantum Energy Research Centre posted two preprints to arXiv claiming they had synthesized LK-99: a superconductor that worked at room temperature and ordinary atmospheric pressure. Not at four degrees above absolute zero. Not crushed inside a diamond anvil at pressures found near the center of the Earth. Just — a thing you could bake in a tube furnace, sitting on a table, letting electricity run through it forever without losing a joule to friction.

If that were true, it would rank among the largest events in the history of materials science. Power grids that don't leak. MRI machines without a liquid-helium habit. Levitating trains as municipal infrastructure rather than a Japanese flex. Fusion reactors demoted from perpetual forty-years-away to a procurement problem.

By mid-August, multiple labs had tried to replicate it and the answer was no.

Here is the part that is actually beautiful, and I want to be careful with it, because it looks like a failure and isn't one.

The tilting in the video was real. It just wasn't superconductivity. A superconductor expels magnetic fields entirely — the Meissner effect — and floats stably, fully, indifferently, like it has opinions about magnets. The LK-99 samples wobbled and hung off one edge, which is what a lump of ordinary ferromagnetic impurity does. The dramatic resistivity drop was real too, and traced to copper sulfide contamination, which undergoes a structural phase transition around 385 kelvin and drops its resistance sharply for entirely boring reasons. When groups at Peking University, Beihang University, the Chinese Academy of Sciences, and the Max Planck Institute for Solid State Research made cleaner samples, the magic evaporated. Max Planck's single crystals came out transparent purple and insulating. The University of Maryland's Condensed Matter Theory Center — which had been live-tweeting the saga with visible enjoyment — called it: game over.

Three weeks. From worldwide claim to worldwide verdict in three weeks. Historically, this kind of thing takes years.

The obvious moral is that science worked and the internet didn't. I believed that when I started writing this. It's wrong, and the evidence against it is sitting inside the story.

Because look at what actually did the checking. Replication attempts were livestreamed. Maryland's theory center — the group that pronounced the verdict — ran its analysis in public, in real time, on the same platform carrying the hype, and enjoyed itself doing it. Prediction markets took positions on grainy footage of a levitating fleck, which is a checking incentive wearing a casino costume. Graduate students in a dozen countries fired up tube furnaces over a weekend because the entire world was watching and there was glory in being the one who settled it.

Historically this takes years precisely because, historically, nobody cares enough to check fast. The attention was the accelerant for the claim and the refutation. They were not opposed forces. They were the same machine running in both directions.

The disorder didn't stay downstream, either. The two preprints bypassed peer review entirely, and they came out of a lab that was not itself in agreement: one was posted by a co-author without the consent of the others, who filed their own version hours later with his name removed. A third researcher told New Scientist the first paper had "many defects" and had gone up without his permission. Korea University's research integrity committee ended up involved. Some of the noise originated at the source.

So "science did its entire job and the network let us down" doesn't survive contact with the case. The honest version is narrower and more useful: the network is a single organ for hype and for checking, and what decides which one wins is whether the claim is cheaply falsifiable.

Which is where the real problem is, and it isn't the one everybody drew from August 2023.

LK-99 was the most favorable possible conditions for verification. Crisply falsifiable — either the resistance goes to zero or it doesn't. Cheap — a tube furnace and a modest pile of materials, within reach of a well-funded undergraduate. Physically reproducible in a weekend. And glamorous enough that thousands of people actively wanted to be the one holding the answer.

Three weeks was not science's ordinary speed. Three weeks was the floor — what's achievable only when every variable breaks right at once.

Now list what doesn't have those properties. Drug efficacy. Nutrition claims. Policy outcomes. AI capability benchmarks. Anything where the test costs millions, or takes a decade, or produces a probability distribution instead of a yes. There is no glory in checking those and there is no tube furnace. The claim goes out at the speed of desire and the correction arrives, if it arrives, to an audience that has already moved.

That's what the episode actually demonstrates, and almost nobody took it away: not that verification works, but how narrow and unusual the conditions have to be for verification to outrun desire at all. LK-99 is the exception we've been reading as reassurance.

It's worth being precise about what the desire was for, because it wasn't superconductivity. Room-temperature superconductivity is shaped almost exactly like absolution: a single material discovery that quietly solves energy, climate, and transport without requiring anyone to change anything about how they live. Not a better technology — a pardon. That's why it moved at the speed it did. We weren't excited about electron pairing. We were excited about keeping everything and giving up nothing.

And when the pardon was withdrawn, a lot of people experienced it as scientists let us down. That construction deserves an agent, so: nobody manufactured that disappointment on purpose. It was manufactured by every one of us who converted a preprint into a conviction inside of an afternoon — myself very much included — because holding a position feels like participating and waiting feels like nothing at all. The labs published a result. The rest of us had already spent it.

The universe, for its part, remains unbothered. It has no obligation to hand us a room-temperature superconductor, and it may simply not contain one. Electrons pairing up and gliding through a crystal lattice without resistance is already an absurd thing for matter to do — a macroscopic quantum effect you can hold in your hand — and the universe gave it to us grudgingly, at temperatures where air is a liquid.

We wanted it to be free. It wasn't. We found out in three weeks, under the best conditions we will ever get, and mostly we learned the wrong lesson from it.

Still — for a species of jumped-up cosmic debris, being wrong out loud and corrected in twenty-one days is not the worst showing.