Eight times in one book somebody says this is the supreme dharma. Eight different answers.
Four of the eight are spoken by a character who is lying at that exact moment.
Not lying about the dharma — the sentences are fine, some of them are the best in the book. Lying about who they are, or what they want, or why they're in the room. The manufactured heretic saying non-violence is supreme dharma while executing a commission to destroy three cities. The messenger making an unanswerable argument on behalf of a demon king who sent him to buy time.
And the ranking never resolves. Chapter 205 has a god state one absolute; chapter 218 has demons state a different one; chapter 274 convenes a formal adjudication and produces a third. Nothing reconciles them, and nothing needs to, because in this book a claim is disposed of by settling its speaker rather than by answering it.
The reflex we all have, and mostly can't see, of routing a sentence through its author before deciding what it's worth.
It's not a bias in the pejorative sense. It's efficient, it's usually right, and there is no time to evaluate every claim on the merits. Attribution is a compression algorithm for judgement, and it works.
The essay is about what it costs, and the Purāṇa is the specimen because it runs the algorithm to failure.
⚑ Whoever wrote that book handed a figure he wanted destroyed the two most humane sentences in it — all those with bodies are equal, and a sharp objection to inherited caste — and never wrote a rebuttal. Not a weak one. None. In fifteen chapters.
That's not carelessness. That's confidence. He knew that assigning a sentence to a discredited mouth was sufficient, and he was right, and it has held for a thousand years.
Here's the move that makes this an essay and not an observation.
The same book shows what actually works, twice, and it isn't argument.
Two arguments win in the whole volume. Gaṇeśa wins one — not by reasoning better, but by putting the Vedas' own truth-value at stake, so that the opponent cannot reject him without rejecting himself. Ghuṣmā wins the other, and closes the book, by hostaging Śiva's own prior teaching back at him.
Neither wins on content. Both win by relocating the claim into a mouth the opponent has already endorsed.
And chapter 304 supplies the control that makes this near-decisive: the same woman's moral appeal fails, and her citation wins, four lines apart.
So the essay's payload isn't stop dismissing people by their source. That's a lecture and nobody takes it. It's the operational version:
⚑ If you want a claim to land, stop looking for better evidence and start looking for the version your reader has already agreed to. That is a completely different research job — and one almost nobody does, because sourcing feels like the work and reframing feels like a trick.
Because it turns on the writer, and it should be written that way.
I built forty-one pages this month with an inline footnote under every claim. That's right. It's saved me repeatedly. It is also, precisely, the infrastructure that makes speaker-based dismissal free — every sentence one click from who said that, in a corpus where a good share of the sources are practitioner transcripts and popular authors and influence manuals, each already carrying a warning tag that doubles as a discrediting index.
The essay should admit that in the first person and not resolve it, because it doesn't resolve.
Filed from the Shiva Purana Volume 2 DEEP INGEST, 2026-08-11. Source pages: The Supreme Dharmas, and Who Says Them and What Makes an Argument Win. The vault-facing half is in A True Sentence in a Discredited Mouth. Log it in META/what-worked.md if it reaches The Platform.