Ask this volume where knowledge sits — is jñāna the thing that gets you there, or a distraction from the thing that does?
It answers four times. The answers do not agree, three of them are the same speaker, and three of the four are inside eight chapters.
⚑ This page does not resolve them. It sets them out, refuses the standard repair, and states what the volume itself never does: this text does not rank its own instruments consistently.
The first is the harshest, and it comes from the creator god, about his own kind.
Those who bypass devotion for knowledge alone
"…are those who pound empty husks from which the grain has been taken out. Nothing but hardship remains for them."1
⚑ Not less effective. Empty. The grain has already gone; the labour is real and the yield is nothing. And it is Brahmā saying it, at 10(14), about the path his own constituency is famous for.
The second is not about jñāna specifically. It is about everything, and jñāna is on the list.
At 11(12), Kṛṣṇa enumerates the instruments — yoga, sāṅkhya, dharma, study, austerity, renunciation, sacrifice, vows, tīrthas, yama and niyama, and knowledge — and states that none of them wins him the way company does.2
⚑ The entire technical apparatus of the tradition, listed by name, ranked beneath sitting near the right people.
⚠ Note the shape of this one. It is not a claim that knowledge is worthless. It is a claim that the whole category of technique is outranked, and knowledge is simply inside the category.
Two chapters on, the same speaker to the same listener:
"What is achieved by strong devotion towards me is not achieved by yoga, samkhya, dharma, studying, austerities or renunciation."3
⚑ Not achieved. Not achieved more slowly, not achieved partially. The list is again the apparatus, and the claim is now exclusion rather than ranking.
And then at 11(19), from the same mouth, seven chapters after position two:
"I love a person who possesses jnana the most. A person who possesses jnana nurtures me. Austerities, tirthas, meditating, donations and other sacred acts do not obtain as much of success as a little bit of jnana ensures."4
⚑ Read it against position one. Brahmā said the jñāna-seeker pounds empty husks. Kṛṣṇa says he loves that person the most.
Read it against position two. There, knowledge sat in a list that lost to company. Here, a little bit of it beats austerities, tīrthas, meditation and donations.
Read it against position three. There, devotion achieved what studying could not. Here, knowledge achieves what sacred acts cannot.
And the same chapter supplies a means/end collapse that makes the ranking question stranger still: "It is accepted that I am the goal and also the means of achieving that goal."
Somebody you trust tells you four times what matters most and gives four different answers. Before deciding which one they meant, check whether the answers were given to different people, at different stages, for different purposes — and if they were not, stop looking for the meaning and record the inconsistency.
You are ranking your own tools. Notice that the ranking changes depending on what you are arguing against, and that this is not dishonesty. It is what ranking is for.
Somebody harmonises two contradictory statements by adding a qualifier neither of them contains. That is the most common way a tradition loses information about itself.
The goal and the means are the same thing. If that is true, every ranking of means is a ranking of the goal against itself, which may be why none of them holds.
There is a standard repair, it is elegant, and it is not available.
The repair: position four is not about knowledge in general but about jñāna informed by bhakti — knowledge that has devotion inside it. On that reading Brahmā is condemning dry knowledge, Kṛṣṇa at 11(19) is praising devoted knowledge, and everything harmonises.
⚑ The text supplies no such qualifier at the point of claim.
Look at what 11(19) actually says. A little bit of jnana. Set against austerities, tirthas, meditating, donations — which are themselves devotional acts. The comparison is knowledge against devotional practice, and knowledge wins. A reading that inserts devotion into the winning term makes the sentence compare devotion to devotion, which is not a sentence anybody would write.
⚠ And the vault's own rule closes it. Adding a qualifier the source does not contain, in order to make two passages agree, is manufactured reconciliation. CLAUDE.md forbids it, and the volume's own narrator is the reason why: at 10(77) he attacks ṛṣis who "do not remember what they themselves have said and are contradictory in their statements." Smoothing this would be doing on the text's behalf exactly what the text convicts others of failing to do.
All four stand.
Though they disagree about the ranking, the four are not random, and the pattern in them is worth stating because it explains the disagreement without dissolving it.
Every one of the four is a demotion of something else.
Brahmā demotes knowledge-without-devotion. 11(12) demotes the whole technical apparatus beneath company. 11(14) demotes yoga, sāṅkhya, study and austerity beneath devotion. 11(19) demotes austerities, tīrthas, meditation and donations beneath knowledge.
⚑ Each statement is made against a rival, and the rival is different each time. What stays constant is the form — one thing raised by naming a list it beats — and what varies is which item is holding the microphone.
⚠ That is an observation about rhetorical shape, not a reconciliation. It does not tell you where jñāna sits. It tells you that this volume ranks by contrast, and that a text which ranks by contrast will produce incompatible rankings if it argues often enough — which this one does, across two Skandhas, to one listener.
And the listener never asks.
Uddhava is present for positions two, three and four. He is the addressee of all three, and he asks questions constantly in this Skandha — about tattva-counts, about intention, about what a devotee does, about the definitions of thirty moral terms.
⚑ He never asks about this.
Nor does Parīkṣit, who is present for the whole recitation, who elsewhere checks the reciter against a version he already holds, and who twice presses on the morality of the rāsa.
⚠ The volume's two auditors are both in the room, both demonstrably willing to object, and neither raises the one question the material most obviously invites. Bounded: no such objection found in the range read.
That silence is worth weighing against what the volume normally does with an objection, because its habit is the opposite. It prints the difficulty first and answers it second. Parīkṣit's chronology query at 10(12) halts the narration and gets a doctrinal reply. His two objections to the rāsa are printed in full and answered two different ways. His "that is what we have heard" at 10(75) forces an explanation of a discrepancy. The apparatus for raising exactly this kind of problem exists, is used repeatedly, and is idle here.
⚑ Which leaves two readings and the volume supports neither over the other. Either the incompatibility is not visible from inside the conversation — each ranking arriving as an answer to a different question, with nobody holding all four at once — or it is visible and is not felt as a problem, because a ranking in this domain is understood by everyone present as a way of demoting a rival rather than as a claim about a fixed order.
⚠ The second reading is the one this page's evidence actually supports, since all four share that rhetorical shape. It is still a reading. The text says nothing, and a reader holding the printed volume — able to turn from 11(12) to 11(19) in a minute — is in a position no participant in the scene occupies.
The strongest support is the speaker-identity. Positions two, three and four are Kṛṣṇa's, to Uddhava, inside eight chapters. Different speakers across two Skandhas would be a compositional observation; the same speaker to the same listener in one conversation is not.
A second support. The lists overlap. Austerities appears in positions two, three and four. Tīrthas in two and four. The same items are being ranked differently against different winners, which is what makes the incompatibility checkable rather than interpretive.
The tension, preserved. ⚑ This is the tension itself; there is nothing further to preserve it against. The page's whole content is that four rankings exist, that they exclude one another, and that the volume never sets them side by side.
A second tension. I am the goal and also the means sits in the same chapter as position four. ⚠ If means and end are identical, ranking means is incoherent in principle — and the volume states both the identity and the rankings without noticing.
Open question. Does the corpus anywhere stage a comparison between two of these four? Not found in the range read — every one of them is delivered as though the others had not been said.
Set against The Narrator Stops the Story to Say That What He Just Narrated Cannot Be True, the volume's own charge is the frame for this page. ⚑ The narrator accuses transmitters of not remembering what they themselves have said, and then supplies four rankings that do exactly that — with the added difficulty that these are not transmitters' statements but the protagonist's.
Against God Gets Up Before Dawn and Does the Whole Thing Anyway, there is a fifth position that is not stated at all. The daily regimen puts meditation on the ātman first, before any rite, and enacts a ranking by sequence. ⚠ It matches none of the four cleanly, and the volume argues four ways and lives a fifth.
Against The Wives Who Walked Out, position one has a narrative counterpart. Ritual competence buying nothing at 10(23) is Brahmā's empty-husks claim in story form, and the two are never connected by the text.
Against He Promised to Explain All Twenty-Four and Explained Twenty-One, the avadhūta's closing principle is the nearest thing the volume has to an explanation: "Jnana obtained from a single preceptor is not stable and may not even be complete." ⚠ That licenses multiple sources; it does not license one source ranking the same items four ways.
Hierarchy of Influence: The Four Levers — influence factors ordered by strength, so a practitioner knows which lever to reach for first and what to fall back on. The hierarchy's value is that it is fixed: a ranking that changes with the occasion is not a ranking.
The Uddhava-gītā is full of hierarchies and none of them is stable. Company over the apparatus. Devotion over yoga and sāṅkhya. Knowledge over austerities and tīrthas. Each is delivered with the confidence of a fixed ordering and none survives contact with the next.
⚑ What neither field gives alone: a practitioner's hierarchy has to be stable because somebody will act on it, and a soteriology's does not, because nobody can check the outcome. The four levers can be tested — you reach for one, and it works or it does not. Nothing in this volume's rankings is falsifiable by the person following them, and that is precisely the condition under which incompatible orderings can coexist for centuries without anybody noticing.
Which is why the inconsistency is worth recording rather than explaining. ⚠ It is not evidence that the text is careless. It is evidence about what a ranking is for in a domain with no feedback — and the volume's own answer, visible in the shape of all four, is that a ranking is a way of demoting a rival, not a way of choosing a tool.
Alter Your Style to the Person — the same content is pitched differently to different people, because what persuades depends on who is listening. Adaptation is competence, not inconsistency.
That is the obvious defence of these four passages, and it fails on a fact the page has already established.
⚑ Positions two, three and four have the same listener. Uddhava, in one conversation, over eight chapters. There is nobody to adapt to.
⚠ Position one is Brahmā at 10(14), a different speaker to a different audience, and the adaptation defence covers it comfortably. It does not reach the other three, and the other three are the ones that exclude each other most directly.
Which sharpens what the defence would have to become. Not different audiences but different moments in one audience's development — a staged curriculum in which each ranking is true for the stage it is given at. ⚠ The volume supplies no stages, no markers, and no indication that Uddhava has moved between chapters 12 and 19. That reading is available to a commentator; it is not in the text, and this page does not adopt it.
Sharpest implication. Brahmā says knowledge-seekers pound empty husks. Kṛṣṇa demotes the entire technical apparatus, knowledge included, beneath company. Two chapters later he says devotion achieves what yoga, sāṅkhya, study and austerity cannot. Seven chapters after that he says he loves the possessor of knowledge most, and that a little of it beats austerities, tīrthas, meditation and donations. Three of the four are the same speaker, to the same listener, inside eight chapters — and neither of the volume's two auditors, both of whom object to other things, asks about it.
Generative questions.