Nārada takes Kṛṣṇa aside after the horse-demon is killed and delivers what looks like a hymn. Partway through, it stops being a hymn.
"O lord! Day after tomorrow, during the day, I will see that Chanura, Mushtika and other wrestlers, the elephant and Kamsa have been killed by you. After that, you will kill Shankha, Yavana, Mura and Naraka. After that, you will defeat Indra and steal the parijata. After that, with the characteristics of viryashulka, you will marry the daughters of the valiant ones."1
And it keeps going. Nṛga freed from his curse. The Syamantaka jewel and a wife. The brahmana's dead son restored. Pauṇḍraka killed. Kāśī burned. Dantavakra and the king of Chedi slain at the great sacrifice.
Then the last item:
"I will see you as Arjuna's charioteer, slaughtering the akshouhinis."1
Every item lands.
Cāṇūra and Muṣṭika, 10(44). Kaṃsa, 10(44). Kālayavana, 10(51). Naraka, 10(59). The pārijāta and the fight with Indra, 10(59). The vīryaśulka marriages, 10(58). Nṛga, 10(64). Syamantaka and Jāmbavatī, 10(56)–10(57). The preceptor's son restored, 10(45). Pauṇḍraka, 10(66). Kāśī burned, 10(66). Śiśupāla at the sacrifice, 10(74). Dantavakra, 10(78).
⚑ Thirteen forward promises, made in chapter 37, and the volume keeps every one of them across the next forty-one chapters, in substantially the stated order.
It is worth being precise, because prophecy is the wrong word for what it does.
A prophecy in a narrative usually creates suspense or dread — something is coming and the characters cannot avoid it. This does neither. Nobody acts on it. Kṛṣṇa does not respond. It changes nothing in the plot.
What it does is orient the reader. Hand someone this speech at chapter 37 and they have the shape of the remaining Skandha: the enemies in order, the marriages, the jewel, the two sacrifices, and the exit into the Mahābhārata.
⚑ It is a finding-aid, printed inside the story, spoken by a character to another character who does not need it.
And the framing makes that explicit. Nārada does not say this will happen. He says "I will see" — I will see that Chanura, Mushtika and other wrestlers… have been killed by you, and at the end, I will see you as Arjuna's charioteer.
The verb is about the speaker's future viewing, not the addressee's future acting. ⚑ He is describing himself as an audience, which is precisely what he is: the corpus's habitual witness, present at events across every Purāṇa, whose function is to have been there.
So the speech is a spectator announcing his programme. That is a stranger and more accurate description of what a Purāṇic prophecy does than prediction, and this one says so in its own grammar.
And this is why it matters beyond its own chapter.
The vault's Volume 2 source stub records that this text's long-range cross-reference system closes — seven forward promises, six paid off inside the volume — and concludes that it is a text that keeps its appointments.
Volume 3 supplies the strongest instance of that system working. It also supplies the clearest instance of it failing: endnote 673 cross-references Chapter 10(86) as though present, and 10(86) is not printed. The front matter lists it at 59 ślokas. The body runs 10(84) → 10(87).
⚠ Bounded, and the bound is not negotiable: absent from the EPUB as ingested. A production defect leaves an identical signature and cannot be distinguished from a recension difference using this file. No claim about Debroy's Sanskrit or the print edition.
⚑ Same apparatus, both outcomes, one book — an index that holds across forty-one chapters, and a cross-reference pointing at nothing.
One qualification, because the claim should be checkable rather than impressive.
The sequence holds substantially but not perfectly. Nārada lists the pārijāta theft and the defeat of Indra before the vīryaśulka marriages; in the book, Rukmiṇī's abduction at 10(52)–10(53) and Nagnajit's bulls at 10(58) precede the Indra episode at 10(59). The preceptor's son is restored at 10(45), which is earlier than several items listed before it.
⚑ So it is an accurate inventory with an approximate sequence — every item lands, and the ordering is thematic rather than strictly chronological: enemies grouped, then acquisitions, then the two sacrifices, then the exit.
⚠ Recorded because a structural claim that overstates its own precision is worth less than one that states its limits. Thirteen of thirteen on content; the order is indicative.
Someone lays out the whole plan at the start, in order, and you half-listen because it sounds like enthusiasm. Two years later you find the note and every item happened in sequence, and the person who wrote it was not predicting. They were describing what they were going to do.
A document contains a reference to a section that does not exist. Everything else in it is meticulous. That combination tells you more about how the document was assembled than either fact alone.
You are handed a summary before the thing itself and it spoils nothing, because what you needed was not surprise. You needed to know which of the next four hundred pages were load-bearing.
The omissions are as informative as the inclusions, and there are three worth naming.
No Vraja. Nothing in the index mentions the gopīs, the rāsa, the departure or the unkept promise of return. The entire separation cycle — sixty chapters of it, and the material the tradition remembers best — is absent from the volume's own summary of itself.
No Uddhava-gītā. The Eleventh Skandha's thirty chapters of doctrine do not appear. The list ends at the Mahābhārata.
No death. Nārada says he will see Kṛṣṇa as Arjuna's charioteer and stops. The destruction of the Yadus at 11(30)–11(31) and Kṛṣṇa's own departure are not in it.
⚑ The index covers the killings, the marriages and the acquisitions. It covers nothing that happens to the women, nothing doctrinal, and not the ending. What the volume forecasts about itself is a campaign record.
This is not a one-off habit. At 12(12) the volume supplies a retrospective summary of itself, and its omissions run the same way — the crimes are kept and the women's grief is dropped.
[n=2 SUGGESTIVE] on the volume's self-descriptions being campaign records at both ends, forecast and recap. ⚠ Two instances in one volume; no claim of a general habit across the corpus, which has not been surveyed for this.
The strongest support is checkability. This is not an interpretation — the thirteen items are named, and each can be located by chapter. Anyone with the volume can verify it in an afternoon, which is rare for a structural claim about a Purāṇa.
The tension, preserved. The volume's cross-reference apparatus is simultaneously the most reliable and the most broken thing about it. T1 stands: the characterisation "a text that keeps its appointments" is correct for Volume 2 read alone and needs bounding, not correcting.
A second tension. The forecast omits everything the tradition most values about this Skandha. Either the volume does not consider the Vraja material part of its plot, or the index is Nārada's interest rather than the text's — and the text supplies no way to choose.
Open question. The endnotes record three further disagreements with Skandha 9, in Volume 2, all genealogical (notes 503, 512, 1307). ⚑ The system keeps its appointments forward and disagrees with itself backward — recorded on the source stub, and the Volume 2 stub should be enriched rather than corrected, since its finding stands for the range it covered.
Set against The Devotee Who Started the Killing, the two chapters are one character running the same faculty in opposite directions, and the volume places them adjacent without remark. At 10(36) Nārada supplies accurate information to an enemy and it produces a massacre plan. At 10(37) he supplies accurate information to a friend and it produces nothing at all, because the friend already knows.
⚑ The same omniscience is decisive in one chapter and inert in the next, and the difference is entirely in who is listening.
Against the vault's Volume 2 material on this corpus's cross-reference discipline, this page is the enrichment rather than the correction. The system does work. Forty-one chapters is a long appointment to keep, and the volume keeps it thirteen times in one speech.
Apocryphal Birth-Prophecy Architecture — the structural role of the prophesied-greatness scene in a life-story: a boy in the dirt, a passing holy figure, a prediction. The prophecy is not evidence about the future; it is a device that tells the audience how to read everything that follows.
Nārada's speech is that device relocated. It is not at the birth, and it is not about greatness in general. It is a manifest, forty-one chapters long, delivered at the exact hinge where the pastoral material ends and the campaign begins.
⚑ What neither field gives alone: a birth-prophecy licenses a reading, and an itemised forward index constrains one. The apocryphal-prophecy form works precisely because it is vague — this child will be great survives any subsequent biography and can be attached retrospectively. Nārada's list cannot. It names Nṛga, the Syamantaka jewel, the pārijāta theft; it is falsifiable by a reader with a table of contents.
That is a much riskier compositional move and it buys something different. A vague prophecy authorises the story; a specific manifest asserts that the story was planned. And it makes the two material lacunae in this volume legible as failures rather than as gaps — because a text that demonstrably keeps thirteen appointments has established the standard by which a missing chapter is a broken one.
Operational Transparency — the open-kitchen effect: showing the process, rather than only the result, increases the audience's satisfaction and their valuation of the work.
The forward index is that move applied to narrative. The volume opens its kitchen — here is what I am going to do, in order, for the next forty-one chapters — and gives up every conventional benefit of withholding.
⚑ The insight: transparency about process is a claim of competence, and it only pays if the process delivers. The open kitchen works because the cooking is good; a visible kitchen with bad practice is worse than an opaque one. Nārada's manifest is the same bet at maximum stakes — the volume publishes its remaining plot and is thereafter auditable against it.
And the audit result is the interesting part: thirteen of thirteen on the promises it makes in this speech, and a failure elsewhere in the same book's apparatus. Operational transparency did not prevent the defect; it made the defect findable. That is the cost the open-kitchen literature does not price — visibility does not improve the process, it improves detection, and a mostly-reliable system that publishes its own standard is more exposed than an opaque one that never did.
Sharpest implication. A character stops the story to read out the next forty-one chapters in order, and the volume delivers all thirteen items — while the same book's apparatus points at a chapter it does not contain. The index covers killings, marriages and acquisitions, and omits the gopīs, the doctrine and the ending.
Generative questions.