This one started as a challenge to something I’d written. In an earlier essay I’d argued that there’s a kind of knowledge — the felt quality of an experience, what it’s actually like from the inside — that can’t be transmitted in words no matter how good the words are; you can describe the colour red to someone who has never seen it in exhaustive physical detail and they still won’t know it the way you do. And the obvious objection is: then what on earth are the wisdom traditions doing? Zen masters spend lifetimes provoking insight in their students with little riddles, and it apparently works. Doesn’t that prove insight can be passed from one mind to another after all?
I think the answer is no, and that working out why it’s no is more interesting than the objection — because it turns out Zen agrees with me, has agreed for about a thousand years, and built its entire method on the agreement.
Note
On how these get written: I read the source material and think it through, a good deal of it in a long back-and-forth with an LLM, and then I lean on the LLM as an editor to turn that thinking into something readable, because I’ve never written much and I haven’t got the patience to grind my own sentences into shape. The ideas, the connections, and the judgments are mine, the wrong ones included. The prose has had help. I’d rather say that plainly than pretend otherwise.
A koan doesn’t transmit anything
Take the famous ones. What is the sound of one hand clapping? What was your original face before your parents were born? Or Zhaozhou’s Mu — a monk asks whether a dog has Buddha-nature and the master simply answers “Mu,” no. These are not puzzles with answers you could write on a card and hand over. A student can sit with one for months. And the thing that eventually happens — if it happens — is not that they are told the answer and understand it. It’s that the question grinds their ordinary problem-solving machinery to a halt, and in the stall something reorganizes, and a realization arises in them that was never in the koan to begin with.
That’s the key move, and it dissolves the objection cleanly: a koan doesn’t transmit content, it triggers a state. It carries no payload of insight, because it can’t, for exactly the reason that you can’t put red into a sentence. What it does instead is act as a perturbation — a deliberately unsolvable demand that exhausts the discursive, concept-shuffling mind until it gives up, and in that giving-up the student’s own system produces the experience directly, first-person. The teacher cannot put the realization into the student’s head. The teacher can only build the conditions under which the student’s head produces it itself. That is self-generation, externally provoked. It is not transmission, and calling it transmission is the whole confusion.
Zen says so itself
The striking thing is that this isn’t my clever reframing imposed on Zen from outside — it’s Zen’s explicit self-description, and it’s almost aggressively clear about it. The tradition characterizes itself with a set of phrases that translate roughly as not founded on words and letters, a special transmission outside the scriptures, and directly pointing at the mind. The most famous image is the finger pointing at the moon: the teaching is the finger, the realization is the moon, and the entire danger — the thing students are warned about constantly — is mistaking the finger for the moon, mistaking the words for the thing the words are aimed at.
So Zen is not a counterexample to “the felt thing can’t be transmitted.” It is the technology built around that limit. A whole tradition organized on the premise that the essential thing cannot be said, only pointed at, and that the pointing only succeeds if the student’s own nature does the rest. They conceded the point I was arguing roughly a millennium before anyone framed it as the hard problem of consciousness.
Three things this clarifies
Once you see the koan as a trigger rather than a transmission, a few sharper distinctions fall out, and they connect to some unexpected places.
A koan is the exact inverse of a new word. When you genuinely need to teach someone a concept they lack, you coin a term and teach it by examples — this is how expertise gets extended, how a chess engine can hand a grandmaster a genuinely new idea about a position, how a model might one day teach us its own internal concepts by naming them. That’s additive: you’re building a new structure in the learner’s mind. A koan does the opposite. It’s subtractive — it dissolves concepts, strips away the conceptual scaffolding, to point at something that lies past the model rather than adding another room to it. Teaching-by-neologism and teaching-by-koan are pointing in opposite directions: one hands you a better map, the other tries to get you to look up from the map entirely.
A koan deliberately sabotages the thing that makes us clever. Our signature cognitive trick is meta-representation — we can model our own models, hold a thought and label it as a thought, doubt our own beliefs. It’s the root of science, of language, of doubt itself. But Zen’s wager is that this very faculty, the endless modeling of models, is also a veil: it keeps you forever in mediation, dealing with representations of experience rather than experience, and you cannot think your way out of it because thinking is the trap. A koan is anti-meta-representation technology. It is engineered to defeat the smartest part of you, because the smartest part of you is what’s in the way.
It’s suspension of disbelief, run backwards. You know the feeling of being so absorbed in a film that the screen vanishes and you’re simply in it — and you know the jolt when a phone goes off and you’re snapped back out, suddenly looking at a construction you can’t un-see. That jolt is a frame-break: it drops you out of immersion into cold, mediated distance. A koan engineers a frame-break too, but aimed the other way — it’s designed to drop you out of conceptual mediation and into unmediated immediacy. Same mechanism, opposite vector. One break ejects you from the experience into the description; the other ejects you from the description into the experience.
India got there first — by a long way
What convinced me this is a real cognitive technology rather than a Zen idiosyncrasy is that the same move runs all through the Indian traditions, and runs deeper there, because it is far older. It’s worth being clear about the chronology, because I had it backwards at first: the koan as a device — the paradox-question formalized into a training system — is a late crystallization, mostly Song-dynasty Chinese Chan, only about a thousand years old. But the move the koan embodies — subtractive pointing, self-inquiry, the dissolving of the ego-self toward a non-conceptual realization — is sitting in the Upaniṣads, which predate the Buddha himself and predate Zen by well over a millennium. Neti neti and tat tvam asi are older than Buddhism, let alone Chan.
And these aren’t independent inventions, which makes the convergence more interesting rather than less. Zen descends, through Mahayana Buddhism, from the very Indian contemplative ferment the Upaniṣads came out of, and Buddhism took its own shape partly in argument with the Upaniṣadic doctrine of the Self. This is one sprawling family working the same terrain over millennia, not strangers stumbling onto the same trick.
The closest single cousin to a koan is the practice of self-inquiry, ātma-vichāra — holding one question, “Who am I?”, and tracing the sense of “I” back toward its source — taught in its sharpest modern form by Ramana Maharshi in the twentieth century, though the impulse is as old as the Upaniṣadic question of the Self. The point is emphatically not to reach a verbal answer; it’s to dissolve the questioner. Ramana said the resolution of the inquiry is the disappearance of the one who was asking. That’s a koan in everything but name — an unanswerable-in-words question that, held long enough, short-circuits the very faculty trying to answer it. And it’s aimed at precisely the target this whole series keeps circling: the self that feels like a fixed thing at the centre of experience. “Who am I?” is a practice for catching that self in the act of being constructed.
The subtractive move — the inverse-of-a-neologism we just described — is in fact older and more formal in the Indian setting. Neti neti, “not this, not this,” from the Upaniṣads, is a disciplined negation of every identification in turn — not the body, not the senses, not the mind, not the thoughts — until what can’t be negated, because it’s the one doing the looking, is left standing. Advaita Vedānta even has a name for the finger-pointing-at-the-moon principle itself: adhyāropa-apavāda, “deliberate superimposition followed by retraction” — you teach with provisional concepts and then withdraw them, because the concepts were always a ladder to be kicked away. The compressed pointing-statements called mahāvākyas — “tat tvam asi,” thou art that — work the koan’s way too: a teacher hands you one not as a proposition to believe but as something to be realized first-person. And the tantric Vijñāna Bhairava offers a whole catalogue of triggers that are pure frame-break-into-immediacy — attend to the gap between two breaths, the seam between waking and sleep, the pause between two thoughts — each one a doorway through a momentary crack in the conceptual stream. Even the destination has a name that gives the game away: nirvikalpa samādhi, literally samādhi without vikalpa, without conceptual construction — the non-meta-representational immediacy, labelled.
The differences are real and worth keeping. Zen, especially in its koan-heavy Rinzai form, leans on the sudden shock of paradox; most of the Hindu paths are more gradual and cultivational, and where they do go sudden — Ramana’s inquiry, the tantric gaps — they tend to defeat the discursive mind by recursion and negation (trace the “I” to its root, negate every predicate) rather than by the head-on absurdity a koan uses. Same lock, different picks.
But the deepest difference is the one I find most revealing, because it cuts the other way. Zen and Buddhism point at anattā, no-self; Advaita points at Ātman, the true Self, indeed the only thing ultimately real. Verbally those are opposite destinations — and, as noted, not arrived at independently: Buddhism sharpened anattā in direct argument with the Upaniṣadic Self it grew up beside, so this is an old family quarrel, not a coincidence of strangers. Yet the method and the reported phenomenology — dissolve the ordinary ego-self, arrive at a non-conceptual immediacy — converge almost completely. Which suggests something that bears directly on everything else here: if the realized state is genuinely ineffable, then “no-self” and “true Self” may be opposite verbal overlays laid across one wordless event — the labels two neighbouring lineages reached for when the experience itself couldn’t hand them the word. It’s the ineffability problem showing up at the scale of whole traditions: feuding cousins describing the same shift, in good faith, with contradictory nouns.
What can and can’t cross the gap
So the refined picture is this. Pointers, methods, concepts, instructions — all of that is transmissible; you can be told how to sit, what to do with your attention, which question to hold. That part is ordinary teaching. But the realization itself — the felt shift — is not transmissible; it can only be self-generated. What a koan adds to “you must generate it yourself” is that the generation can be triggered: a well-aimed perturbation can substantially raise the odds that your system reorganizes.
And the trigger is unreliable and non-algorithmic by its nature, which is not a flaw in the method but a consequence of what it’s doing. You cannot koan a rock into enlightenment. The substrate has to already have the latent capacity for the state to arise; the koan doesn’t supply the experience, it only occasions it, so it only works on a system that could already, in principle, have had it.
Which leads to a genuinely strange thought, and the reason this sits on a blog that’s half about AI. Could you koan a large language model? Pose it the unsolvable question and see what happens? The trouble is that a system with no phenomenal inside would do exactly what a system with a phenomenal inside might also do under the surface — produce fluent, plausible, enlightenment-flavoured text. If there’s nothing it’s like to be the model, the koan triggers no state, only more output; if there were something it’s like, the koan might genuinely reorganize it — and from the outside you could not tell the two cases apart, because the only evidence you’d ever get is more language, and language is precisely what the koan exists to point past. The koan turns out to be a probe aimed exactly at the thing our instruments can’t reach. Which is, when you sit with it, its own small koan.