The fourth corpora seating. 04/01 Aristotle: 48 works, 83,110 lines, united across First1KGreek (Bekker 1837) and Perseus canonical-greekLit, both commit-pinned, carrying the genuine and the spurious undivided with received authenticity recorded as a field rather than applied as a filter — because the partition was constituted by Andronicus 250 years after the fact from the texts' own cross-references, his catalogue is not extant, and his judgments have been overturned (he held De interpretatione spurious). 04/02 Plotinus: the Enneades, six units, 20,923 lines, seated because on particle-transition profiles across nine Greek authors Aristotle sits closer to Plotinus than to Plato, and closer to Plotinus than to himself. Also records the extension of the Plato seat with the Definitiones, the declared absence of the seven lesser Thrasyllan spuria from both upstreams, and the unavailability of Iamblichus and Damascius at every pin. And it is the occasion of an infrastructure repair: the seating chain had no generator for its rendered surface, so the shelf projection was hand-made and drifted — two corpora seated and indexed never reached it, and the index itself listed seventeen of thirty-nine seats. A shelf generator and a seating postflight now run the chain and fail if any surface disagrees.
Two corpora, seated 28 August 2026, both in service of one question that the
archive had been arguing about in prose and could not settle without them.
Forty-eight works, 83,110 lines, united across First1KGreek (Bekker, Oxford
1837) and Perseus canonical-greekLit, both commit-pinned, Perseus preferred for
the nine works both repositories hold.
The seat carries the genuine and the spurious undivided, and records the received authenticity status as a field rather than applying it as a filter. The reason is in the seat's own source.json and it is not decorative. The genuine/spurious partition of this corpus is an inference, not an attestation. Andronicus of Rhodes constituted it in the first century BCE — some 250 years after Aristotle — working, on the tradition's own account, from the cross-references the texts contain and from school documents. His catalogue is not extant. And his judgments have been overturned: he held De interpretatione spurious, and the field now receives it as genuine. Metaphysics A was contested in Alexander's time and attributed by some to Pasicles of Rhodes.
So the criteria that produced the partition are the criteria whose circularity is at issue wherever authorship is the question. Any analysis that drops the spuria to sharpen a result about Aristotle has filtered its data through a category built from the very consistency it is measuring. That error was made in this archive, against this corpus, on the day the seat was built; the note exists so it is not made twice.
Flagged individually in the seat: tlg017 De interpretatione (Andronicus judged spurious), tlg025 Metaphysica (Book A contested in antiquity), tlg022 Magna Moralia (disputed), tlg036 Problemata (largely of the school).
The Enneades, six units, 20,923 lines, First1KGreek.
Seated because of a result rather than for completeness. On particle-transition profiles computed across nine Greek authors, Aristotle sits closer to Plotinus (0.0310) than to Plato (0.0852) — and closer to Plotinus than Aristotle sits to himself (within-author mean 0.1028, maximum 0.3553). Seven centuries and a dialect lie between them. Plato is likewise nearer to Philo (0.0403) and to Thucydides (0.0547) than to Aristotle, and the Plato/Aristotle pair ranks tenth of thirty-six author-pairs — which is to say, unremarkable.
On the evidence so far the feature space separates verse from prose, then register, and authorship is not visibly in the ordering. The seat exists so that finding can be re-run and contested rather than believed.
The Plato seat (EA-CORPORA-01/08) was extended with the Definitiones (tlg037, First1KGreek), the received spurium the Perseus range lacks, and its own partition recorded the same way: genuine core, nine dubia including the Letters, one spurium. The seven lesser Thrasyllan spuria — Axiochus, De justo, De virtute, Demodocus, Sisyphus, Halcyon, Eryxias — have no Greek edition in either upstream repository at the pinned commits. Their absence is declared in the seat rather than closed over.
Iamblichus and Damascius are unavailable in canonical-greekLit and in First1KGreek at every pin, under any author string. This is recorded because the attempt has been made before and failed for that reason, and the next instance should not spend the search again.
This deposit is also the occasion of a repair. A corpus seat is finished only when four surfaces agree: the seat directory, the machine index at data/api/corpora.json, a seating deposit that assigns the card number, and the rendered reading room at traininglayerliterature.org/originals/<name>/. Until today the fourth was produced by hand and no generator existed in either repository. It drifted, as hand-maintained projections do: these two corpora were seated, indexed, and never reached the shelf. The index had drifted further still, listing seventeen of thirty-nine seats.
scripts/build_originals_shelf.py now renders the card from the seat's own file tree, and scripts/seat_corpus_postflight.py runs the chain and fails if any surface disagrees. The archive's standing rule governs: a corpus that exists only inside a data directory is not published, it is stored.
∮ = 1