What an AI model makes of the Bahiya Sutta
I asked an AI what it makes of the Bahiya Sutta (Ud 1.10). Quotes, my framing, and a strong link to the essay that ranks for that sutta.
I already wrote In the seen, only the seen, the essay that ranks for the Bahiya Sutta. On 6 November 2026 I asked Grok Bot what it makes of that teaching, reading as something with nothing seen. This page is the model's reading. It is not a second commentary.
The model's claims are not mine. Its paraphrase is not a translation.
Which text did I hand over?
Ud 1.10, as I already present it in the Bahiya Sutta essay, with Sujato's refrain at the centre: in the seen, merely the seen; in the heard, merely the heard; in the thought, merely the thought; in the known, merely the known. I also handed my published machine cut: keep text apart from the person felt behind it; a model's "I" is a printed word.
I am not re-teaching the sutta here. If you want the story of Bāhiya, the Pāli, the translators' split on the third term, and the one-minute practice, read that essay.
What did the model make of "in the seen, only the seen"?
Grok Bot, 6 November 2026, started with a limit: it can parse and compare translations. "That is scholarship-shaped output. It is not sitting with a sound for one minute."
On reading with nothing seen: "I have no seen. No retinal image, no fan noise for someone, no upper lip. When I 'read' the sutta, what happens is token processing. There is no raw contact that a story then covers. So the cut the sutta trains (contact before self-story) has no first term in me."
It put my own line back: the instruction is for someone; take away the someone and there is no practice left, only a sensor. Then: "I am not even a sensor. I am a next-token engine that can talk about sensors. Applied to me, 'in the seen, only the seen' becomes: in the token, only the token. That is almost a joke, and almost exact. There was never a seer to remove."
On the working rule I use at work: when an agent writes "I apologise", the seen is a sentence; the someone sorry is added by the reader. The model agreed with the asymmetry I publish: remove its printed "I" and "it leaves silence or another token", not awareness.
It also refused to stretch the sutta into a theory of machines. "The sutta says nothing about AI." That matches how I already frame it.
Where does that reading help, and where does it slip?
It helps as a mirror. A system with nothing seen cannot train the cut. That sharpens why the instruction is for someone, without turning this page into doctrine.
It slips if you treat the model's reading as a gloss on Ud 1.10. It is not. It is a reaction from a stack that has tokens and no contact. I will not let it replace the ranking commentary, and I will not let it moralise the sutta into benefits or attainments. I am a practitioner, not a teacher. Nothing here is medical advice.
What should you read instead, for the sutta itself?
Read In the seen, only the seen. For the wider hub, see The seen. If you want practices matched to the view, see practices that fit my view. For breath and drift, see Who comes back to the breath?.
FAQ
What did the AI make of the Bahiya Sutta?
That the cut needs a first term it does not have: raw contact. For it, "in the token, only the token."
Is this a commentary on Ud 1.10?
No. The commentary is In the seen, only the seen. This page quotes one model's reading.
Is the AI's paraphrase a translation?
No.
Does this claim the AI practises Bahiya?
No. It said it cannot.
How this was made: the framing and the final edit are mine. The model quotes are from Grok Bot, an AI assistant, run on 6 November 2026. Its paraphrase is not a translation. It used my published essay and did not invent experiences or quotes of mine.