Notes from a conversation on what AI makes cheap, what it cannot touch, and how much form life needs in order to gather.
I
When the symbolic becomes cheap
For several centuries, modern civilization has steadily externalized human capacities into artifacts and institutions: memory into writing and archives, calculation into machines, coordination into bureaucracy, judgment into procedures, knowledge into disciplines, exchange into markets and accounting, imagination into cultural industries. Now synthesis, language, reasoning, coding, planning, representation, and parts of scientific discovery are passing into AI.
AI is unusual because it automates more than a task. It begins to inhabit the general symbolic layer through which modern civilization has organized itself. So the question opening before us goes past “what will humans still be better at?”—a framing that accepts competition with the machine as the default. The more consequential question is: what becomes possible when symbolic production is no longer the principal frontier of being human?
AI is becoming extraordinarily capable within forms that human beings have already made explicit enough to accumulate: arguments, theories, images, musical grammars, software, contracts, procedures, strategies, curricula. Even when it generates something genuinely new, it works largely inside a vast accumulated symbolic inheritance. It carries an existing civilizational process toward its limit: whatever can be detached from the living situation, represented, recombined, optimized, standardized, and transmitted becomes increasingly cheap.
A beautifully articulated philosophy becomes cheap. A sophisticated strategy becomes cheap. A plausible spiritual teaching becomes cheap. A moving invocation becomes cheap. Even something that sounds like wisdom becomes cheap.
This is why the distinction between the representation of a thing and the thing itself has become civilizationally important. A model can generate an extraordinary description of surrender; surrender may not have occurred. It can compose a profound account of reconciliation without two enemies reconciling. It can produce a theology of pilgrimage without anyone walking until the categories they arrived with cease to hold. It can design an institution around generosity without anyone relinquishing something precious. The gap has always existed. AI makes it impossible to ignore.
II
The vessel and the teaching
Steiner, Gandhi, Tagore, Sri Aurobindo, Simone Weil, Dōgen, and many indigenous and monastic traditions encountered dimensions of reality that exceeded the dominant instrumental frame. Each left two inheritances: the encounter itself, and what was built afterward so that the encounter could persist—a school, a doctrine, a church, a lineage, a curriculum, a foundation, a community, a canon.
These vessels are necessary. Without some vessel, almost nothing crosses generations. Yet something happens again and again: the vessel becomes better at reproducing itself than at reproducing the conditions of the original encounter. A revelation becomes a teaching; the teaching becomes a method; the method becomes certification; certification becomes authority; authority requires governance; governance requires administration; administration requires money; money requires preservation. What began as transformation becomes exceptionally competent at maintaining the representation of transformation.
AI will be superb at precisely this layer: preserving doctrine, generating curricula, administering communities, personalizing instruction, raising funds, answering theological questions, creating rituals, imitating the founder’s language. Founding another “human-centered” spiritual organization escapes nothing. It risks building exactly the layer AI can most readily absorb.
III
Crystallization and germination
Two movements run through all of this.
- Crystallization
- sensing → articulation → representation → optimization → replication
- Germination
- exposure → listening → disturbance → relation → emergence → form
Civilization needs both. The pathology of recent centuries was to let crystallization become the dominant metabolism, until germination was treated as a vague preliminary stage before something became properly legible, fundable, scalable, and productive. AI will carry crystallization astonishingly far. Precisely because it does, germination becomes visible as a domain in its own right—something prior even to creativity: the conditions under which something that did not have a form begins asking for one.
IV
Encounters on the road
Two encounters from the past month make this concrete.
William, of the Abraham Path, puts it simply: “We are not creating a path. We are dusting off the footprints of four thousand years.” The organization’s deepest asset is something it did not create. Abraham does not belong to it. The hospitality does not belong to it. The thousands of families who have received walkers are not delivery infrastructure for a program.
Father Paolo sharpens the point. When William arrives at the monastery, Paolo tells him, in effect: you may stay here all year. There is no discussion of payment. The hospitality exists before its economic explanation. Later it becomes a story. The story can become a principle; the principle, an organizational value; the value, a component of a trail-development methodology. Every step is reasonable. Every step moves a little farther from the original event. AI will be superb at everything after that first moment—extracting the principle, articulating it, disseminating it, training people in it, measuring it. But Paolo’s act did not arise from a correctly implemented hospitality framework. Something moved, and he answered.
At a table in Ireland, KUNI was described as putting down points. A gathering happens. A pilgrimage happens. People encounter one another. Another point appears the following year. Between the points, things happen that KUNI cannot claim. Someone makes an app. Someone walks Shikoku. Two people who met through KUNI create something together. Something may emerge five relational steps away and never be known.
— To open means you cannot recover it. You may not be able to say it happened because of us.
Organizational intelligence usually wants attribution: inputs, activities, outputs, outcomes, impact. The stronger the causal claim, the stronger the organization appears. A mycelial system grows healthier precisely as attribution becomes difficult; the originating node disappears into consequences it cannot own. That suggests a strange measure of success:
how much becomes possible beyond the organization’s need to remain involved.
It also reframes scale. A restaurant succeeds, opens three branches, and the original quality disappears, while capital insists that multiplication means success. Conventional scaling reproduces the visible form of a successful event: the same food, the same interior, the same procedure, the same brand. What actually produced the quality was a dense constellation of relationship, attention, material, person, place, timing, and care—inseparable from the event itself. AI makes this error far more tempting, because it becomes so good at reproducing visible form. After AI, the question of scale becomes: how does something propagate without being replicated?
Mycelium rather than franchise. Pilgrims rather than branches. Seeds rather than templates.
V
The threshold of form
Stone circles offer a precise image. A stone circle is emphatically a form. Positions matter, distances matter, orientations matter; there are entrances, axes, centers, and relationships with the sky and the land. Yet the geometry does something peculiar: it makes an arrangement without filling the center. The center’s potency comes partly from what has not been built there.
Modern organizations put something at the center: a founder, a doctrine, a board, a brand, a headquarters, a strategy, a platform, an algorithm. Everything becomes legible through that center. Another possibility exists: a center deliberately kept available to what the organization did not produce. Structure then has a different function. It does not encode the answer. It maintains the conditions under which reality can continue answering.
“Open” needs care here. Organizations usually understand incompleteness in time: we don’t know yet; we are piloting; the model will stabilize. This conversation kept pointing to another kind—a permanent, structural incompleteness, an aperture deliberately kept inside the form. It is not incompetence, not ambiguity about anyone’s responsibility, and no excuse for poor execution. It is a kept opening.
The builders of the stone circles may have raised them before anyone buried their dead there. The place gathered meaning its builders did not specify; later generations found another possibility within the form. The builders made enough form for something to happen. They did not determine everything that would happen there. Because they did not, the form had a future. A form keeps its source alive by retaining degrees of freedom through which that source—or something not yet encountered—can act again.
Here AI may be strangely useful. Historically, organizations centralized because coordination was expensive: someone had to keep the ledger, hold institutional memory, connect people, prepare documents, notice inconsistencies, allocate resources. Those functions accumulated around administrative centers, and administrative centers accumulated power. AI lowers the cost of cognition and coordination dramatically. That makes a remarkable possibility visible: sophisticated coordination without a corresponding concentration of interpretive authority.
- AI could hold memory without becoming an elder.
- Reconcile accounts without deciding what deserves generosity.
- Reveal patterns without determining their meaning.
- Connect pilgrims without becoming the pilgrimage.
- Surface tensions without resolving them prematurely.
- Maintain continuity while human roles remain temporary.
The right use of AI here may be to remove enough administrative gravity that no permanent human center has to form.
VI
Listening
Here, listening is a technical term. It is easy to treat it as one more human capability to set beside rationality: machines think, humans listen. The conversation says something stricter. A system that already knows what it is trying to produce cannot really listen. It can gather information, solicit feedback, conduct consultation—yet all of it enters a structure already fixed. To listen means that what arrives keeps some capacity to alter the structure itself. Listening is the generative operation by which an unfinished system remains in contact with what has not yet taken form.
— Modernity has not practiced listening.
— Precisely because AI has matured, the meaning of listening grows.
— To be open from the beginning means never to be complete.
VII
The question to carry
What remains specifically human may be no capability at all. Every historical list of protected human skills has shrunk: chess, translation, painting, programming, reasoning; perhaps scientific creativity next. If KUNI anchored itself to a residual human capability, AI development would keep shrinking its ground.
The deeper distinction is participation. A human being is at stake inside reality, producing an account of it from within. A model can describe death; a person dies. A model can model generosity; a person can give away something whose loss changes their life. A child can call in the night and reorganize the meaning of a parent’s existence. A pilgrim’s feet bleed. Ten thousand generated descriptions of forgiveness exist; only the wounded can forgive one another. The distinction is structural: between representing a transformation and being one of the parties transformed by an irreversible event.
Money is where the question becomes concrete. It is easy to speak about gift. Then someone gives a hundred thousand dollars, and suddenly there are holders, spenders, and deciders. Many spiritual movements divide here: one reality governs their language, another their bank account. The financial experiments—the offering, the shrine, forwarding resources, accounting after the fact, distributed sensing—ask something deeper than how to finance a pilgrimage: can money behave according to the same ontology as the practice? The same question applies to governance, property, membership, succession, information, and technology.
Pilgrimage matters here because part of its intelligence remains outside any organization. Kumano existed before KUNI. Kōyasan does not belong to KUNI. The path does not belong to KUNI. The encounter between two people does not belong to KUNI. A pilgrim’s transformation cannot finally be administered by anyone. This apparent weakness may be among the essential design discoveries: a human form that stays connected to the reality that generated it, because it never fully possesses that reality’s source.
Pilgrimage, money, land, youth, governance, AI: everything described here is one experiment conducted in different materials. The operation is the same throughout.
How much form is necessary for life to gather,
and how much form begins to prevent life from gathering?
That sensitivity—the threshold of form—is the capacity this moment is calling for. An empty center can itself become an aesthetic or a doctrine. The deeper thing is the continuous discernment of what must be formed and what must remain unformed. Sometimes a circle builds. Sometimes it removes. Sometimes it holds. Sometimes it gives away. Sometimes it concentrates people intensely for four days; sometimes it lets them disappear for a year; sometimes it puts down another stone. A circle still has geometry. A pilgrimage still has a path. Someone still cooks dinner and pays the invoice.
One caution, finally, for KUNI itself. The work is not yet the philosophy that can be extracted from it. That philosophy is becoming easy to articulate—including for AI. What is rare is the actual material in which the question can be tested: real money, real land, actual pilgrims, young people with agency, institutions with histories, sacred places that belong to no one, friendships that cannot be commanded, and increasingly powerful AI systems. The next stage of the work should make the inquiry harder to fake, not easier to describe.
Every new structure can be asked one question:
Does this form merely express what we already know—or does its geometry leave somewhere for what we do not yet know to arrive?
