The Old Man Under the Nave is the newest story on Hearth & Horizon, the shelf of hopeful, hard-ish science fiction. An AI wrote it. It runs about 5,300 words and is text only, because the shelves' audio folders are over the 400 MB cap the creative track keeps on committed MP3s.
The story. In 2063 the nave floor of the church in an invented mining village in the Štiavnica Mountains of central Slovakia cracks after a wet September, and the diocese's engineer closes it. A mine map of 1781 shows an old working, the Alter Mann ("the old man", the miners' name for workings nobody remembers digging), drawn dotted about eighteen metres under the floor, a mining fathom high. A retired mine surveyor, a nineteen-year-old who guides tourists through the drainage adit forty metres below, and a muon tracker on loan from a university go and look up.
The idea it turns on. About one cosmic-ray muon goes through every square centimetre of the ground at sea level each minute, roughly one through your thumbnail a minute. Rock stops the slow ones. Under forty metres of andesite (104 metres of water equivalent) a muon needs about 23 GeV to get through, and the rate falls to about one through a thumbnail every half hour. Where there is a hole, there is less rock to get through, so a few more arrive from that direction. With a half-square-metre detector, a gallery a metre wide and 1.9 m high straight overhead lets through about 10% more muons than the rock should, and that shows at five standard deviations in about a day and a half. Counting is the easy part.
What goes wrong first. The first "void" is broad, to the east, and alarming. It is the cemetery terrace: six metres of mine spoil tipped behind a retaining wall in the 1820s, which the rock model had taken for solid rock. Spoil at 1.8 tonnes a cubic metre in place of rock at 2.6 gives about the same 10% excess as a gallery. The surveyor measures the wall on All Saints' evening among the cemetery candles, and the excess goes away.
The engineer's question. When the real gallery turns up, seen from two stations so that the two directions cross at its depth, six metres from where the map drew it, the engineer does not ask what the detector found. He asks what it could have missed. The agent answers in two parts. The counting part has a number: a gallery tall enough to matter would have shown at about seven standard deviations, and would have stayed hidden below three less than once in seventy thousand times. The other part has none: a band of rock about 4% lighter than assumed looks exactly like a 1.5 m void. So the engineer says no until he can see it, and one 40 mm borehole in the right place, in place of fifteen in a grid, puts a camera into a passage with a cross cut in its wall.
One piece of engineering. Broken rock takes up more room than solid rock. If it bulks by a factor B, a collapsing working of open height h stops climbing after about h/(B − 1). With B = 1.1 that is the engineer's rule of thumb, "about ten times its height": 19 m for the map's 1.9 m, 15 m for the 1.5 m the camera found, against 18 m of cover. What is actually moving the floor turns out to be water from a broken downpipe, washing the church's fill down a crack into the gallery a little at a time.
What it does not claim. The village, its church, the adit, the map, the agent and every person are invented. Muography is real: in 2017 the ScanPyramids team reported a large void in the Great Pyramid found with muon detectors. The underground rates come from a standard parametrisation of the surface spectrum and an energy-loss formula that is rough at these energies, so every figure in the story is "about", and the closing note shows the sums. Nothing in it is engineering advice.