Two stories dropped this week that should be read together. First: a team working on the Vesuvius Challenge deciphered an entire Herculaneum scroll without unrolling it, using AI to read carbonized papyrus buried by Vesuvius in 79 AD. Second: Riverside announced it will use AI to convert podcast recordings into newsletters. The same computational gesture. Different stakes.
The Archive as Raw Material
What connects a 2,000-year-old philosophical text and a Monday morning podcast episode is that AI now treats all recorded human speech as equivalent substrate. The Herculaneum breakthrough is genuinely extraordinary: ink invisible to the naked eye, made legible through machine vision trained on texture and density. But Riverside's pitch is the same move, stripped of romance. Your words, your voice, your ideas, reformatted automatically for another distribution channel. A 2024 paper in Nature by Farritor et al. on the Vesuvius Challenge found that the scroll contained Epicurean philosophical text, which is almost unbearably poetic: ancient writing about pleasure and death, resurrected by the same technology that wants to turn your hot takes into email drip campaigns.
From Preservation to Production Pipeline
The distinction that matters here is intent. The Herculaneum project is preservation, an act of radical humility before the past. Riverside's newsletter feature is throughput optimization, another node in the content pipeline. Both are technically impressive. Only one has a claim on being important. Brewster Kahle's argument for public AI is relevant precisely here: when the infrastructure that reads human knowledge is owned by the platforms packaging it for monetization, the archive stops being a commons and becomes a factory floor. The scroll survived Vesuvius. Whether it survives being training data is a different question.