Week five. The teacher has just shown the four-headed ram form of Banebdjedet and asked why a divinity of the thyroid would need four heads.
He pulls up a parathyroid diagram. Four small glands, behind the one. Number matching number.
It is the strongest claim in twelve hours of material.
Then, minutes later, in the same session, on the same image:
"These four heads are the Ba of Ra, Ba of Shu, Ba of Geb, Ba of Asar, all connected in the same body."
Four spirits riding in one vessel, with a car analogy — you're driving, four people get in, they contribute direction.
He does not reconcile them. He does not mark one as figurative. Both stand.
[RESONANCE]
Because I had spent three clusters grading this method's outputs — strongest instance here, weakest there, this one constrained, that one free — and had been treating variation in quality as the thing to track.
This is variation in number. Given one input, the method produced two outputs, and the operator didn't notice.
Which is a different kind of fact. A method that produces variable-quality answers is a method that needs a grader. A method that produces two incompatible answers to the same question, without flagging it, is a generator — and generators produce output whether or not there is anything there.
The vault already holds this shape in the si / jo / men cluster comparison: three roots, three completely different internal structures — one collapsing to an image, one spreading and needing a connector supplied, one containing a proposition.1
I read that as informative about the roots.
It may be informative about the reader. A method flexible enough to find one motion in nine senses, and two pairs in four senses, and an argument in eight, is flexible enough to find something in anything.
And the corpus supplies the corroboration itself: it never reports a failed reading. Fifteen divinities, six triads, and every single one yielded a siting.2
No misses. Not one.
Against the corpus's own critique of invented cosmology — the Anuket memes, made up to sound deep, circulated, adopted as identity.3
The teacher's distinction would be that his readings are derived and theirs are invented. Which is real, and is a difference in process.
But a derivation procedure with no stopping rule, no self-test, and a zero failure rate is behaviourally indistinguishable from invention with better inputs.
Promotion criterion: find one case, anywhere in the Odwirafo corpus, where a reading was attempted and abandoned. One published failure would break this. Its absence across a very large corpus is the evidence.
Search the wider corpus — the Ḥonam Su, Okra Okraa and Akradinbosom courses already in the vault — for any instance of an abandoned identification. The vault holds enough of this teacher's material to run the check without new sources.