You are on the night shift at the Pierrelatte enrichment plant in southern France, June 1972. Your job is the dullest in nuclear science: check that uranium is uranium. Every rock on Earth, every meteorite, every grain of sand contains exactly 0.7202% uranium‑235. It is a constant. It is boring. Tonight the machine says 0.7171 — and somewhere in Gabon, a rock is about to admit it ran a nuclear reactor for a few hundred thousand years, two billion years before anybody thought of one.
Uranium comes in two main flavours. Uranium‑238 is the heavy, sluggish majority. Uranium‑235 is the rare one — the one that splits. In every uranium sample humans had ever measured, anywhere on Earth or off it, U‑235 made up 0.7202% of the atoms. Not roughly. Exactly. You could set a watch by it, and the whole enrichment industry did.
So when the mass spectrometer at Pierrelatte reported 0.7171% in a batch of uranium hexafluoride, nobody shouted "discovery." They assumed the machine was dirty. They cleaned it. They ran it again. Same answer. Then they walked the number backwards through the whole chain — the gas, the oxide, the yellowcake, the ore — and the trail led to one mine, in one country: Oklo, in Gabon.
By the time they had audited the paperwork, more than 200 kilograms of uranium‑235 were missing from shipments. Some ore samples came in at 0.64%. One came in at 0.44%. That is not a dirty instrument. That is uranium that has been used.
Load a sample into the mass spectrometer. The pen draws a peak for each isotope, and the counter prints the U‑235 fraction. Six samples are ordinary. One is not. Find it — then run the two follow‑up tests the French chemists ran when they refused to believe their own machine.
Here is the hinge the whole story swings on. Both uraniums are radioactive, but they decay at wildly different speeds. U‑235 has a half‑life of 704 million years. U‑238 has a half‑life of 4.468 billion years — more than six times longer. So the rare isotope is disappearing six times faster than the common one, and every year that passes, natural uranium gets a little weaker.
Run the film backwards and the opposite happens: uranium gets richer. Turn the dial below and watch a boring grey rock turn into something a power station would pay for. The curve is not a cartoon — the page is computing it live from the two half‑lives.
Stand on this shore and there is nothing to stand on but rock. No grass, no trees, no insects, no soil. Rivers run brown and unbraided by roots. The only life worth mentioning is in the water: mats of cyanobacteria, quietly doing the most consequential thing that has ever happened on this planet — breathing out oxygen.
That matters here for a reason nobody would guess. Uranium does not dissolve in water unless there is oxygen around. In an airless world, uranium stays locked in the mineral it was born in, scattered thin, useless. Once oxygen appears, water can pick uranium up, carry it for kilometres, and then dump it all in one place the moment it meets something that strips the oxygen back off — a bed of black, organic‑rich mud.
So the Great Oxidation Event, roughly 2.4 to 2 billion years ago, did something odd and wonderful. Bacteria that had never heard of physics started concentrating nuclear fuel. In the sandstone of the Francevillian basin, uranium piled up until parts of the ore body were 20 to 60% pure uranium — and the seams grew tens of centimetres thick, metres across.
Two clocks had to line up for this to work, and they only overlapped once. Uranium‑235 had to still be above about 3% — which stopped being true a long time ago. And oxygen had to have already arrived — which had only just become true. Oklo sits in the narrow window where both were briefly, simultaneously, correct.
A fission chain reaction needs one thing above all: slow neutrons. A neutron flying out of a splitting atom is far too fast to be caught by the next one — it screams straight past. Something has to slow it down. In a modern power station that job is done by water, and it is also how the station is controlled.
At Oklo the water was groundwater, seeping through the pores of the sandstone. And that gave the rock something no engineer would dare design: a reactor whose moderator boils away when it gets too hot, and comes back when it cools down. Below is that machine. Neutrons are real objects in this simulation — they are born fast, they slow down when they hit water, they get eaten by poisons, and sometimes they split an atom and make more.
Claiming a rock ran in half‑hour pulses two billion years ago sounds unfalsifiable. It isn't — and the proof is one of the loveliest pieces of detective work in geochemistry. In 2004, Alex Meshik and colleagues at Washington University found xenon gas trapped inside tiny grains of aluminium phosphate that had grown around the uranium: the highest concentration of xenon ever measured in any natural material.
Fission does not make xenon directly. It makes short‑lived iodine and tellurium first, and those decay into xenon a while later. Each xenon isotope has its own precursor with its own half‑life — some decay in seconds, some take days. So whether a given xenon atom was born in the rock or born in the water that had already washed away depends entirely on how long the water stayed gone.
Set the dormant window below and watch which xenon isotopes end up in the cage.
Oklo made about six tonnes' worth of fission — and on the order of two tonnes of plutonium, all of which has long since decayed away. Every argument about burying nuclear waste eventually asks the same question: will it stay put? Oklo is the only full‑scale, two‑billion‑year experiment anybody will ever get.
Guess before you look. For each element, decide whether it is still sitting in the rock where it was made, or whether it escaped.
Seven questions. Instant feedback, no penalty for guessing, and the machine is not judging you — it has been wrong before.