Picture a road through scrub forest, and lying across it, something the size of a hill.
The boys stop. They look at it. And then — this is the part nobody remembers about this story — they take it apart, piece by piece, out loud.
The upper lip, one of them says, looks like a cloud gone red at sunset. The lower one on the ground looks like that cloud's shadow. The corners of the mouth are caves in a mountainside. The fangs are peaks. The tongue is a broad road. Inside it's as dark as the gap between two summits. The breath coming out is a forest fire's wind.
And then two more observations, and these are the ones that should have ended it:
"There is the bad stench of burnt corpses. There is a smell of flesh emerging from the inside."1
They can smell the dead. They say so. They are children with cowherds' pouches and sticks and no weapons, standing in front of a thing that smells of eaten people, and they have described it so accurately that a modern reader could draw it from their words alone.
Then they walk in.
Here's what did not happen. They were not fooled. They did not mistake the snake for a hill. The narrator gives us the first impression — some of them thought it was a pretty spot in Vrindavana, snake-shaped, a place to play in — and then he lets them talk, and the talk demolishes that impression in nine consecutive clauses. Perception is fine. Description is fine. The nose is working.
What breaks is the last step, and the text prints it as a question:
"If we enter, will it devour us? Or will it be instantly destroyed by this one, like Baka?"1
This one is Krishna, standing next to them. Baka is the last enormous thing that tried this and died. So the inference is not stupid. It runs: the danger is real and here is its inventory; there is also a backstop, and the backstop has a demonstrated record; therefore the expected outcome of entering is a dead asura, not dead boys.
Then: "Glancing towards the face of Baka's enemy, they laughed, clapped their hands and entered."
They look at him first. They laugh. Nobody in this scene is confused about anything except the one thing that matters.
Śuka closes the sequence with a flat judgement: "They were ignorant about the actual truth and thought what was false to be true."1
That is the tradition's summary and it is not wrong, but it is a coarser description than the scene it summarises. Nothing in their sensory report was false. What was false was a conclusion drawn from a true premise about a real protector — and the falseness of it depends entirely on facts they had no access to, namely what Krishna had decided to do and when.
Notice that he had in fact decided. Two sentences later the text tells us he "heard and desiring to protect those who were his own, made up his mind to prevent them" — and they were already inside. The rescue was coming. It arrived late. The boys' reasoning was not wrong about the backstop's existence, only about its latency, and latency is the one variable a track record cannot report.
Then the volume does something it almost never does: it prints the deliberation.
"What should be done now? Should one allow the wicked one to live? Or should one prevent injury to the innocent and the virtuous? How can both be achieved?"2
Read that as a bind and not as a rhetorical flourish. Kill the asura now and the boys inside die with it. Save the boys and the thing survives. This is not good against evil; it is two goods that will not fit through the same door.
And immediately before it, four words that the volume will spend the next sixty-five chapters denying it ever said. He was filled with compassion, "but also surprised at what destiny had done."2 At 10(77) the narrator will interrupt his own story to insist that sorrow, delusion, affection and fear cannot exist in him, and will call the ṛṣis who say otherwise incoherent. That denial does not address this passage. It does not appear to know this passage is there.
You are three years into a job where one person has never once let a project fail. Something lands on your desk that you can see is broken. You can list what's broken about it — you do list it, in a thread, accurately, with timestamps. Then you ship it, because they have never let one go down.
You are on a ridge and the snow has that hollow sound. You say so. Your friend says so too. Neither of you turns around, because the guide is thirty feet ahead and the guide has done this line forty times.
You are reading the second page of a contract and you find the clause. You read it twice. You say the words "that's a problem" out loud in an empty room. Then you sign, because the firm on the other side has a name you have known your whole life.
In none of these is the failure a failure of attention. In each of them you produced the correct inventory and handed it, intact, to a piece of arithmetic about somebody else's reliability. That is the shape. The inventory is not the safeguard. The inventory is what makes you feel safe enough to stop.
The most interesting three seconds in the scene are procedural.
One boy raises it: "O friends! Tell us if it is without life, stationed in front of us. Doesn't it resemble the gaping mouth of a snake that will devour us. Isn't that true?"1
And the answer comes back: "True." — and then the whole nine-clause inventory follows as elaboration.
So there was a check. Somebody asked. The group confirmed. And the confirmation deepened the description rather than dissolving it, which is exactly what a functioning verification procedure looks like. Then the group used its own confirmed finding as the input to the backstop calculation and walked in together, laughing.
⚑ The check ran, returned the right answer, and made the entry more likely rather than less. Having said the danger out loud and heard it agreed to, they had already handled it — that is what saying a thing in company does. The group had performed diligence, and the performance discharged the anxiety that diligence exists to preserve.
A record of past rescues carries two pieces of information and withholds two others.
It tells you the protector is capable. It tells you the protector is willing.
It does not tell you when. And it does not tell you whether this case is inside the set the record was built from.
The boys had both of the first two, honestly earned — Baka was real, and Baka is dead. What they did not have was any way to read the interval between Krishna noticing and Krishna acting, and the text is explicit that an interval existed. He heard, he made up his mind, and "however, they entered inside the asura's stomach."
The word doing the work in that sentence is however. It marks a gap the boys could not have measured from outside.
The Tenth Skandha opens its Vraja sequence here, and the placement is doing work. Everything that follows in the boyhood — the year Brahmā steals and nobody notices, the poisoned pool, the mountain lifted for seven days — runs on the same premise the boys reasoned from. There is a backstop, and it does hold. The text does not undercut that. It simply shows, once, at the front, what it costs to reason from it.
There is a second bounded observation here about how this volume measures harm. The threat's inventory ends on the smell of corpses; the trilemma is about "injury to the innocent"; and elsewhere the volume reaches for children specifically when it wants a unit of cost — the infant left at the breast when the gopīs run to the rāsa (10.29), the infants cast aside in Mathurā (10.41), and cows and women miscarrying from an asura's bellow alone (10.36). [n≥3 PATTERN] on the nursing or unborn child as this text's chosen measure. Here the unit is the whole cohort, and the cohort walks in on its own feet.
The strongest support is that the text supplies the reasoning rather than the outcome. A story that wanted to teach heed danger would have had the boys miss the signs. This one has them win the observation contest and lose anyway, and prints the winning observations at length. That is a deliberate compositional choice, and its cost — a nine-clause description of a monster that nobody acts on — is high enough that it is hard to read as accident.
The tension it leaves open, and the page does not close it. Śuka's verdict is "ignorant about the actual truth" — a verdict about knowledge. The scene shows a failure of inference under a correct premise. Those are different diagnoses and the text supplies only the first. It is tempting to say the volume means the boys were ignorant of Krishna's nature rather than of the snake, and that reading is available; but the text does not say it, and reaching for it would be supplying a reconciliation the source declined to write. Preserve both: the described failure and the assigned name for it do not match.
A second tension, running to 10(77). "Surprised at what destiny had done" sits sixty-five chapters before the narrator's insistence that no such state can arise in him. Neither passage acknowledges the other. This is not evidence that the corpus is confused — it is evidence that the denial at 10(77) is bounded by material it never addresses.
Open question. The rescue was decided before it was executed, and the gap is what killed the arithmetic. Does this volume anywhere else print a delay between the decision to protect and the protection? At 10(16) he holds the position in Kāliya's coils "for a muhurta" deliberately, so that the witnesses arrive and suffer — which is a timed delay, but a chosen one. Whether the Agha gap is the same mechanism or a different one is not determined by the text and is not claimed here.
The vault's Volume 2 material records that contact, not merit, is this corpus's operative variable — the demoness who offers a poisoned breast gets a mother's destination because she offered it. That page and this one watch the same machinery from opposite ends. There, nearness to Krishna converts an intention to murder into an attainment. Here, nearness converts an accurate threat report into a decision to be eaten.
What Volume 3 adds at its first Vraja chapter is that the override runs against the person too, and that the text shows this without softening it.
Set beside Govardhana and the Six-Year-Old Umbrella, this is the harder companion piece. The mountain is protection delivered on time, to people who asked. This is protection delivered late, to people who did not ask because they had already counted on it. Eleven chapters apart, and the volume never puts them side by side.
Attribute Substitution: Trust as Proxy for Competence — Kahneman's mechanism: asked a hard question it can't answer, the mind silently swaps in an easier one and reports that answer instead. In front of a used car you can't evaluate, you answer do I trust this seller — and never notice the swap.
The boys do the opposite and land in the same ditch. No swap happens. They answer the hard question, correctly, in nine clauses, out loud. Then they run a second calculation about the backstop and let it override the first.
Attribute substitution says the substituted judgement is invisible. Here it is spoken, and its premises are true.
⚑ What neither side gives alone: accurate perception is no defence against the trust heuristic, because the heuristic doesn't need to suppress the perception — only to arrive after it. The buyer who does know engines, inspects, finds the fault, and buys anyway on the dealer's forty-year name has not made Kahneman's error. He has made this one, and the psychology material has no name for it.
When the Crowd Is Wrong Together — that page's argument is that social proof exists because verification is expensive, so the audit gets outsourced to the crowd. The failure mode is that nobody actually ran it; everyone was reading everyone.
Here the audit ran. The group did it, on the record, and it came back correct. And they went in anyway — because the confirmation got consumed as reassurance instead of as data.
⚑ The insight neither domain produces alone: a verification ritual that completes successfully can raise the risk of the thing it was meant to gate. Saying the danger aloud and hearing it agreed to discharges the need to act on it. The behavioral literature treats unperformed diligence as the problem. This scene supplies the opposite case, and it is why the boys laugh and clap before they enter. They are not ignoring the check. They are relieved by having done it.
Sharpest implication. The most dangerous configuration is not blindness — it is accurate observation held next to a reliable protector. The record of a thing that has always worked is a premise, and a valid inference from a true premise can still walk you into a mouth, because the premise cannot report its own latency. The boys were right about the snake and right about Krishna and wrong about nothing except when.
Generative questions.