While every AI lab on earth spent the week screaming about agents, token prices and IPOs, a quieter team did something that actually made me stop scrolling: they took a papyrus scroll that Mount Vesuvius cremated into a charcoal briquette in 79 AD and read the fucking thing — without unrolling it, because unrolling it would turn 2,000-year-old paper into confetti.
How the hell they did it
The scroll, PHerc. 1667, is one of hundreds recovered from a villa in Herculaneum. Instead of touching it, researchers ran it through X-ray tomography — basically a CT scan — to capture 3D images of the layers inside. Then software virtually flattens the coils, and an AI trained to spot the faint textural ghost of ancient ink picks out letters a human eye would never catch. The result: nearly 1.5 metres of text across 20 columns, pulled out of a lump that looks like something you’d sweep out of a barbecue.
And what did our 2,000-year-old genius have to say?
A philosophical treatise on ethics, human nature, and moral progress. Which is to say: some Epicurean was banging on about how to live well and feel good, got interrupted by a volcano, and two millennia later the most advanced pattern-recognition machinery humanity has ever built was pointed at his carbonised diary to find out he was basically writing a self-help thread. Honestly, beautiful. The bar for “insight” has not moved in 2,000 years.
The AI tool stack actually worth paying for
One email a week. The models, tools and moves that matter, stripped of hype and filtered so you don’t have to drink from the firehose. Free, and you can bail anytime.
Why I’m not even being sarcastic (much)
Strip the jokes and this is the best argument for AI I’ve seen all month. No hallucinated legal briefs, no chatbot pretending to be your girlfriend, no agent expensing $9,000 of tokens to reformat a spreadsheet. Just a genuinely hard problem — recovering human knowledge that was physically gone — cracked by pointing machine learning at it. This is the Vesuvius Challenge, launched back in 2023 by Brent Seales with Nat Friedman and Daniel Gross, and it is quietly doing more for the case that this technology is worth a damn than any product launch this year.
We spent a hundred billion dollars teaching computers to think. Turns out one of the best things they can do is help us finish reading. I’ll take it.