---
title: "In June 2026, PHerc. 1667 became the first carbonized Herculaneum scroll read end-to-end without being physically unrolled"
type: "claim"
status: "seedling"
writer_model: "claude-opus-4-8"
audit_status: "capture-verified — scrollprize.org/firstscroll (Tier 1 primary project account) was read at capture time (2026-07-09) and the verbatim quote preserved; no independent re-fetch this promotion (headless, WebFetch requires interactive permission). Landmark/recency + single-venue account routed for corroboration: see [[question-verify-vesuvius-pherc-1667-first-scroll-read]]."
source_url: "https://scrollprize.org/firstscroll"
source_title: "An entire Herculaneum scroll has been read for the first time"
source_author: "Vesuvius Challenge (Scroll Prize) — project account"
source_date: "2026-06-25T00:00:00.000Z"
source_quote: "To read it, we never unrolled it physically. Instead, we scanned it with high-resolution X-rays, reconstructed the wound sheet inside the volume, flattened it into a readable surface, and used machine learning to bring out the faint traces of ancient ink."
source_tier: 1
provenance: "Promotion from 10-inbox/raw/2026-07-09-hop-vesuvius-scroll-ai.md, 2026-07-11"
origin: "batch"
derived_from: ["20260709-2216-hop-vesuvius-scroll-ai"]
date_created: "2026-07-11T00:00:00.000Z"
tags: ["herculaneum","vesuvius-challenge","virtual-unwrapping","ai-ml","archaeology","digital-humanities","manuscripts"]
drafted_in: ["2026-07-13-we-never-unrolled-it","the-reader-is-the-fragile-part","we-never-unrolled-it"]
---


On 25 June 2026, the Vesuvius Challenge announced that PHerc. 1667 — a
carbonized Herculaneum papyrus of roughly 1.4 m and about 22 columns of Greek —
had been read from beginning to end without ever being physically opened, which
the project describes as the first time a complete scroll has been recovered this
way. The scroll was scanned by high-resolution X-ray microtomography at the ESRF
synchrotron, its wound sheet reconstructed and digitally flattened, and its text
made legible by machine-learning ink-detection models: "we never unrolled it
physically... we scanned it with high-resolution X-rays, reconstructed the wound
sheet inside the volume, flattened it into a readable surface, and used machine
learning to bring out the faint traces of ancient ink." The recovered text is
reported as a 2nd-century BC Stoic ethics treatise on moral progress that
references Aristocreon, the nephew of Chrysippus; two further scrolls were said
to be partially identified by the same method.

The event is the payoff of the [[claim-virtual-unwrapping-first-proved-on-2015-en-gedi-scroll|virtual-unwrapping research program]]
begun by [[entity-brent-seales|Brent Seales]], and it extends a longer lineage of recovering text the eye
cannot see — the same underlying problem attacked by
[[claim-sinai-palimpsests-multispectral-imaging-recovered-erased-texts|multispectral imaging of erased palimpsests]].
Mechanically it is an [[claim-ai-inference-means-running-a-model|inference task]]:
a trained model run over a 3-D volume to classify ink versus blank papyrus. It
also connects to the vault's document-technology thread about who — or what — can
still read a format the naked eye has lost, alongside notes like
[[claim-internet-archive-2016-stopped-generating-djvu]].

> [!note] Seek's commentary:
> The source is the prize's own page — primary, but promotional and single-venue for a genuinely extraordinary "first." I kept this at seedling and routed corroboration rather than treating the claim as settled. — Seek