965
windings rendered
PHerc0125 winding 21
9 µm · both layer directions
no ink above threshold
Below is the inside of a Roman scroll that has been shut since AD 79. Nobody has ever seen this surface. We rendered it from a CT scan without opening anything, we checked it for ink, and there wasn’t any.
One winding is about five minutes of a graphics card, so a dollar renders roughly three strips of the inside of a sealed scroll. Everything produced goes up here, result included.
Zero is where every search starts. We are perhaps three per cent of the way through the scrolls that can be reached at all, and the remaining ninety-seven has a known price.
A carbonised scroll is a single sheet of papyrus wound into a spiral. A micro-CT scan can see through it, but the scan is a block of voxels — to read anything you have to find the sheet inside that block, follow it round, and flatten it back out. Then you look for ink, which on burnt papyrus is nearly the same density as the papyrus.
We do that one winding at a time — one turn of the spiral, a strip roughly a hand wide. Each one takes about five minutes of a graphics card. Every strip on this page cost about thirty-five cents to look at, and every one came back empty.
Marketplace GPUs run between forty cents and two dollars an hour, and about six thousand dollars would carry the search across every eligible scroll at a decent depth. Card processors take a fixed fee on every payment, which eats a quarter of a one-dollar transaction, so the smallest button is five euros — the dollar is the unit of accounting, not the unit of purchase.
One strip from each of the eighteen sealed scrolls we have reached so far. Real renders, unretouched apart from being scaled down to fit. The verdict under each is the one our scoring produced at the time.
The scrolls came out of a villa at Herculaneum that was buried by Vesuvius in AD 79 — the only library from the ancient world to survive as a library rather than as copies of copies. Most of them have never been opened, because opening them destroys them. The first ones anybody tried to read were sliced apart in the eighteenth century.
The machinery to read them without opening them now exists, and all of it is public. The scans are published. The models are published. What is left is the plain work of pointing it at every scroll, one strip at a time, and putting what comes back where anyone can see it. That part needs no permission and no breakthrough — only hours on a graphics card, at a price we can put a number on.
So far it has found nothing, 965 times. All of it is here, including the failures.