You've seen a family do this.
Somebody outside the family says the wrong thing about the grandfather — gets the story slightly wrong, or the tone — and the room goes cold and the correction is instant and it is not kind. That version is wrong, and everyone present knows exactly how wrong.
Then at Christmas the same family tells four incompatible versions of the same story across one afternoon and nobody blinks. The uncle's version contradicts the aunt's on a matter of fact. Both get told. Both get laughed at. Nothing is corrected.
It isn't hypocrisy and it isn't confusion. The rule was never about the facts. It was about who was speaking.
The second chapter of this book does exactly that, at the scale of a theology, in about forty verses.
Most of Ādi 2 is an argument, and the argument is sharp. The claim under defence is that Kṛṣṇa is svayaṁ bhagavān — the original, not an expansion of anybody. Specifically, not an expansion of Nārāyaṇa, which is what a rival reading holds.1
The tone is not ecumenical. The opponent is told: "you say something contradictory and become angry when this is pointed out."2 That is a debating-hall sentence, and it is aimed.
The proof offered is the candle image. Light one candle from another and the second is not lesser — but one of them was lit first, and the order is a fact even though the flames are equal.3 Which is a genuinely economical way of holding two things the tradition needs: full divinity in the expansions, and an original that the expansions come from.
So far this is a doctrinal chapter doing what doctrinal chapters do. Then it ends.
The closing move gives away the entire field:
"Some say that Śrī Kṛṣṇa is directly Nara-Nārāyaṇa. Others say that He is directly Vāmana. Some say that Kṛṣṇa is the incarnation of Kṣīrodakaśāyī Viṣṇu. None of these statements is impossible; each is as correct as the others."4
Every position the chapter has just spent itself refuting is now correct.
And the reason given is not that the arguments were weak or that the matter is beyond us. It is a criterion about the speaker:
"To call Him Kṣīrodakaśāyī Viṣṇu does not add to His glory. But such words from the lips of a sincere devotee cannot be false."5
The proposition is unchanged. Its truth-value has moved, and what moved it is who said it.
It's tempting to file this as tolerance, and tempting in the other direction to file it as incoherence. Neither fits.
Look at what the concession actually protects. It does not permit the opponent's version — the opponent has been dealt with, at length, and nothing here reopens that. It permits a devotee's version, and only that. The set of sayable things has not expanded; the set of people whose sayings pass has.
Which makes this a boundary rule rather than a doctrinal one. The proposition is load-bearing at the edge of the community and decorative at the centre, and both facts are stated openly, thirty verses apart, in the same chapter, by the same author, without any sense that an explanation is owed.
The corpus does this more than once. It insists on the plain, direct meaning of scripture when refuting Māyāvādins, and then produces eighteen readings of one verse for an audience of one and sixty-one for a friendly assembly.6 Literalism for enemies, multiplicity for allies. Same shape, different register.
The cheap reading is that "a sincere devotee cannot be false" is a licence — say anything, so long as you're inside.
The text doesn't quite support that, and the difference is worth getting right.
The concession is offered about a specific class of claim: which of several exalted identifications applies to Kṛṣṇa. Every one of the permitted alternatives is a form of the same being. Nobody is being licensed to say he was a minor demigod, or a good teacher, or nonexistent. The permitted range is bounded by devotion itself — you can only say the things a devotee would say, and a devotee's errors are all in the direction of praise.
So the rule is something like: within the space of statements that arise from devotion, correctness of content stops mattering. That's much narrower than "anything goes," and it explains why the chapter can hold it alongside the polemic without feeling contradictory from inside.
It also explains what the polemic was for. It was not for establishing a fact about Kṛṣṇa; the fact is negotiable, as the ending admits. It was for establishing the boundary — for making clear that the question is asked by people who are outside, and that being inside means the question stops mattering. The argument's function is the drawing of the line, and the line is what the chapter is about.
There's a test available for that reading, and the chapter passes it. If the polemic were really about the content, the concession would have to be qualified — you would expect something like these alternatives are permissible for beginners, or provided they are understood correctly. Qualifications are how a doctrinal writer protects a claim he still cares about. There are none here. The alternatives are declared as correct as the others, flatly, with the only condition attached to the speaker's sincerity.
A second piece of evidence, from the other direction: the polemic's heat is aimed at people who are arguing, not at people who are wrong. The line about saying something contradictory and becoming angry when it is pointed out is a complaint about conduct in a dispute. It is what you say to an opponent across a table, and it has no bearing at all on whether Kṛṣṇa is an expansion of Nārāyaṇa. The chapter's sharpest sentence is not about the doctrine, and that is the tell.
Two more things in Ādi 2 belong on this page because they are doing the same work in different registers.
The tenth object. The Bhāgavatam's ten subjects, of which nine exist to isolate the tenth: "To distinguish this ultimate shelter from the other nine subjects, the mahājanas have described these nine."7 The scripture's visible content is contrast material for the one thing it is actually about — which is another way of saying that most of what a text argues is scaffolding for a boundary.
Argument as exercise. "A sincere student should not neglect the discussion of such conclusions, considering them controversial, for such discussions strengthen the mind. Thus one's mind becomes attached to Śrī Kṛṣṇa."8
That second one deserves to be set beside a sentence from the far end of the book. After the Junior Haridāsa episode, the corpus instructs the reader: "Do not argue, for arguments will produce a contrary result."9
Argument is a spiritual exercise when it establishes the doctrine and a spiritual danger when it questions the discipline. Both instructions are given as general principles, neither is qualified, and the corpus never places them together.
Layer A runs the polemic and then makes the concession, without transition or apology. Kṛṣṇadāsa is comfortable with a rule that operates differently at the boundary and the centre, and he is comfortable saying both parts aloud.
Layer B does not soften the polemic — if anything the purports sharpen the Māyāvāda material, which is the commentator's standing preoccupation. What it does with the concession is more interesting: it reads it devotionally, as an instance of the tradition's generosity toward sincere feeling, rather than as a rule about who gets adjudicated.10 That's an available reading. It is also the reading that keeps the polemic's force intact while blunting the concession's implication, which is the direction Layer B moves whenever a verse loosens a rule.
You're in an argument about something factual with someone outside your field, and you're winning, and you notice you'd never press this hard against a colleague who said the same thing. Stop and ask what you're actually defending. Often it isn't the claim. It's the line, and the claim is what happens to be lying on it.
Somebody in your own camp says something loose and you let it go, and the same sentence from outside would have got a correction. That asymmetry is not hypocrisy — it's a boundary rule doing its job, and it works exactly as long as nobody says it out loud. Ādi 2 says it out loud, which is the unusual thing about it.
You're evaluating a claim and you catch yourself weighing who made it. Don't pretend you aren't. Instead, get precise about what the speaker's identity is actually licensing: if it's a bounded range of errors that all lean the same way, that's a defensible heuristic. If it's the whole proposition, you've stopped assessing anything.
You're teaching and you notice the disputed questions are the ones that engage people. That's the chapter's own observation — argument strengthens attachment. Now notice what happens when the argument turns on your own practice rather than your doctrine, and whether the same enthusiasm survives it.
The evidence is a single chapter containing both moves at full strength: the polemic with its named opponents and its candle proof, and the concession with its explicit "each is as correct as the others."11 Neither has to be reconstructed.
The tension the source doesn't close is the one this page is about, and the corpus leaves it wide open. A proposition is defended to the death against outsiders and declared optional among insiders, in one chapter, with no bridging argument.
A second one, sharper because it spans the book. Ādi 2: discussion of contested conclusions strengthens the mind and should not be neglected. Antya 2: do not argue, arguments produce a contrary result. The first is said about doctrine, the second about a disciplinary case that ended in a man's death. Nothing in the corpus marks the distinction, and the reader is left with two unqualified general instructions.
Open: is the speaker criterion bounded by devotion, as this page argues, or is that a charitable reconstruction? The text says a sincere devotee's words cannot be false, without qualification. The bounding is my inference from which alternatives are actually permitted, and nothing in the range read states it.
This page is where the vault's Acintya-bhedābheda: The Gauḍīya Position shows its social face. That page holds the metaphysics of simultaneous oneness and difference as a philosophical settlement. Ādi 2 shows the same structure being used as a membership device — the difference is enforced outward, the oneness is available inward — and the two readings support each other uncomfortably well.
Against Advaita Vedānta: Śaṅkara the relation is the one the chapter is picking. Worth noting that the corpus's actual quarrel is narrower than it sounds: Ādi 2's concession shows it will tolerate very large variation in which divine identity is affirmed. What it will not tolerate is the removal of the person. The line is not drawn where the polemic's heat suggests.
And it should be read against Six Diversities in One from the previous chapter, which declares six categories identical. Ādi 1 collapses distinctions by ontology; Ādi 2 collapses them by speaker. Two chapters, two mechanisms for the same permissiveness, and both are stated at the head of a book that will spend sixty chapters drawing distinctions.
Fixed Beliefs vs. Mobile Opinions — Le Bon's observation is that Galileo, Newton and Leibniz could turn full reasoning power on planetary motion and could not turn it on hellfire, because the dogma was not a belief they held but the substrate they thought from. You cannot examine the rock you are standing on.
Ādi 2 does not fit that model, and the misfit is the useful part.
Kṛṣṇadāsa's community can plainly see the belief. They argue it in public, hold rival positions on it, and are told in as many words that the rival positions are all correct. There is no invisibility here at all — the substrate is on the table, being handled. And yet the same proposition is defended against outsiders with a ferocity the substrate model would predict.
So there's a third category between fixed and mobile, and neither page names it: a belief that is rigid at the boundary and negotiable at the centre. Not a substrate — the holders can see it. Not an opinion — outsiders cannot touch it. What it is, structurally, is a shibboleth: a proposition whose content has been retired and whose function is to sort speakers.
That's worth having because shibboleths are easy to mistake for either of the other two. From outside they look like substrate — try to argue and you hit rock. From inside they look like opinion, because everyone in the room holds a slightly different version and nobody minds. Both readings are wrong, and the diagnostic is precisely Ādi 2's: watch what happens to the same sentence in two different mouths. If the truth-value moves with the speaker, you are not looking at a belief at all, and no amount of better evidence will do anything to it.
Le Bon's cases can in principle be dislodged by stepping off the rock. This kind cannot be dislodged by argument in any form, because argument was never what was holding it up.
Sharpest implication. The propositions a community fights hardest over are often the ones it cares least about, and the fight is how the edge gets drawn rather than how the truth gets found. Ādi 2 is remarkable not for doing this — everyone does it — but for saying both halves out loud, thirty verses apart, without embarrassment.
Generative questions.