Grain Rains from Heaven, Ghosts Weep at Night—Huainanzi, “Discourse on the Root”#

“In ancient times, when Cangjie invented writing, grain rained from heaven and ghosts wept in the night.” Writing had scarcely appeared when grain began to fall and the ghosts began to cry. The “Discourse on the Root” in the Huainanzi gives a human invention this earthshaking force, as though the world could no longer settle into its accustomed order once those strokes had been drawn.¹

Read only the rain of grain, and the story readily becomes a hymn to civilization: Cangjie has done a great deed, and heaven and earth celebrate. But why do the ghosts weep? The text goes on to tell of Boyi digging wells, dragons ascending into dark clouds, and spirits retreating to Kunlun. Then comes the judgment: “The more their abilities increased, the thinner their virtue became.” Writing and well-digging gave people greater powers while disturbing an existing order of nature and human affairs. Virtue here belongs with naturalness and a simple way of life. Ability and virtue pull upon each other; increasing power raises a question about how life is ordered.

Today, words on a screen are astonishing us in another way.

A feeling not yet clear enough to explain can be set down in a few sentences and receive a response. A problem whose point of entry is still obscure can be divided, compared, and reorganized. Researchers can use AI to find material and distinguish concepts; engineers can enlist it in derivation and implementation; writers can discover unexpected possibilities through successive rewrites. Where the necessary tools and authorization are in place, language can also connect to operations: after a sentence, a file changes, a program runs, and a new result returns to the conversation.

The first things we notice are usually the visible products. Is the answer clear? Is the essay good? Has the work been done? These judgments matter. Over time, however, people also acquire a way of beginning things, habits of distinguishing good work from bad, and a new understanding of what they can do. A topic once impossible to approach may gradually become their own research. An assessment occasionally borrowed from elsewhere may become a standard by which they judge themselves.

People act through writing, and in acting they change.

This essay asks how language-mediated human–AI practices take shape, differentiate, and change. What do we treat as an object to understand and work on? What demands tell us whether the work is fitting? What sort of people do we become through repeated understanding and action? These questions affect one another. How we grasp an object shapes what we do; what we do changes available capacities and established standards; changed capacities and standards may alter what we expect of ourselves. Research, engineering, writing, learning, and everyday interaction reveal these relations in different ways. Concrete ways of collaborating should grow from understanding them.

The old book preserves an astonishment that convenience has not yet worn smooth. Writing could make heaven and earth change color; how would those who acquired its power go on to live? Today that question follows agents that can answer and act, entering the work on our screens and moving toward the spaces we inhabit.

When Form Is Established, Pattern Is Complete; When Sound Arises, Wen Is Born—Liu Xie, “Tracing the Dao”#

In Liu Xie’s “Tracing the Dao,” the opening chapter of The Literary Mind and the Carving of Dragons, there is already wen in heaven and earth before anyone has composed a work of literature.

The sun and moon display their forms; mountains and rivers spread their patterns. Tigers and leopards bear markings, clouds their own colors. Woods resound, springs and stones give rise to rhythms: sound, too, has an audible organization. “When form is established, pattern is complete; when sound arises, wen is born.” Wen here includes visible forms and colors as well as the organization of sound. When Liu Xie asks how it could be mere external ornament, he reaches beyond the idea of decorating something already complete. Form and pattern together constitute the appearance of things.²

Following this thought into language makes a familiar account inadequate: thought is completed in the mind, and language then expresses it. When checking a paraphrase, we can provisionally distinguish something already said from another way of saying it. Where creation and understanding occur, however, much meaning emerges as a sentence, image, or sound takes shape. A pause in a poem sustains or interrupts a voice; narrative order changes what readers can know at a given moment; placing quantities together in an equation makes a previously unnoticed relation available for reasoning. Form helps determine what we understand. Changing it can alter even the content we thought we already knew.

“Tracing the Dao” moves from the patterns of heaven and earth to human feeling, expression, and cultivation, within an order of the Dao and the sages. Other Daoist texts describe the relation between wen and coming into being in stranger terms. The Record of the Divine Talismans of the Three Caverns speaks of “the naturally soaring, mysterious qi, gathering in the void to form writing,” and of the essential qi of the Dao being spread through writing materials and ink to meet the essential qi of all things. In this account of the world, wen can precede the person holding the brush. A talisman’s configuration, orientation, and material support all participate in its efficacy.³

The same text distinguishes talismans, writing, and diagrams. Talismans draw on the configurations of clouds, things, and stars; writing articulates sounds and sentences to make their purport clear; diagrams depict the forms of numinous transformations. They work in relation to one another, each with its own uses. Understanding an explanation does not give someone possession of the talisman. Recasting a diagram as clear prose may fail to preserve the relations it originally sustained. Verbal intelligibility and the way a written or graphic form exists cannot stand in for each other here.

The preface to the Ancient True-Form Diagrams of the Five Peaks of the Cavern-Mystery Numinous Treasure begins with the configurations of mountains and waters: winding ridges, uneven heights, flowing waves like brush tips, clouds and forests like written characters. Sacred vision discerns names within these configurations. For those who carry the diagrams into the mountains, the text promises that the spirits of the mountains and rivers will come beyond their domains to welcome them. The diagram on paper meets the path through the mountains, changing how heaven and earth receive its bearer.⁴

Yet a diagram does not simply find its way into anyone’s hands. A “Prayer for the Transmission of the Diagrams” leaves places for the date, lineage, and names, to be occupied by particular people giving and receiving it. The transmitter states its provenance and the grounds for passing it on; the recipient expresses a wish to cultivate, protect, and remain faithful to it. The graphic form grants a capacity, and its transmission prescribes a life. What this person can do, in what capacity, and in what continuing relations with others are written into the ritual together.

Graphic forms thus have concrete means of taking effect. Carrying a diagram, entering a mountain, transmission, and vows organize form, bodies, implements, and lineage, placing someone in relations of responsiveness, action, and obligation. Talismans and diagrams acquire their efficacy within this religious world; programs depend on code, equipment, permissions, and arrangements of use. Each requires its own explanation of efficacy. Reading them together draws attention to a shared question: through what conditions does a form acquire power, and into what life does that power bring its bearer?

Xunzi’s “Rectifying Names” makes another way of taking effect explicit. People use their senses to recognize sameness and difference and establish distinct names, enabling them to understand one another, form expectations, and work together. “Names have no fixed appropriateness; agreement establishes what they designate.” Naming admits convention. Yet “names do have a definite goodness: direct, easy, and unobstructive names are called good.” Agreement should make mutual understanding easier. Names are neither simply given by nature nor indifferent in their effects.⁵

The same chapter also speaks of “clarifying noble and base above, distinguishing sameness and difference below.” Names make things distinguishable and assign people their places. To admire the flexibility of convention while passing over this hierarchy would make rectification sound far too innocent. A name can identify a thing clearly and just as clearly install a person in a position they can scarcely change. The correct use of language and the legitimacy of an order therefore require their own arguments.

These materials bring several aspects of language into view at once. Form participates in how things appear; expression enters action; people enter new relations through acquiring a language and its powers. Understood broadly, language leaves room for mathematical notation, images, music, and bodily gesture, each with its own ways of organizing experience. Bodily fatigue, the resistance of wood, and the failure of a device give those ways their measure in use.

This gives linguistic mediation a working meaning: we use discernible forms to understand, communicate, and organize action, and those forms help shape the objects and relations we understand. Mediation names this connected passage. A name makes a difference jointly recognizable; a diagram lets several relations be viewed together; tools turn an instruction into an operation. People have to apprehend and use each form, connecting it to bodies, materials, and institutions. Explaining these connections takes a theory of language into practice.

Not Blinded by What Has Accumulated—Xunzi, “Dispelling Blindness”#

This essay began with a fairly direct question. Since AI can often make language accurate and comprehensive, should we learn to use language in the same way? Research and engineering seem to give us good reasons for pursuing such norms; literature, everyday interaction, and personal experience make other demands. The discussion initially followed a classification of activities and arrived at seven working types.⁶

The classification helped. We began to see how a conjecture and a conclusion, an expression and a promise, could all consist of language while undertaking different things. Later, as the objects, relations of participation, and norms involved in practice became clearer, a separate layer of activities began to look redundant. I asked whether we still needed it. The discussion then turned to understanding concrete practices directly, using classifications where they helped.

What had helped us understand became something we needed to understand.

That was not the last change. After the essay had been written, I asked for its concepts and practical guidance to be made more concrete, with an explicit framework of the kind one might find in a science or engineering paper. The AI added working definitions, a four-part description of practice, conditions for judgment, and examples of handing work over. The result was orderly and contained many sound distinctions. Yet it largely organized the problem around describing tasks, checking their grounds, and moving work forward.

I was dissatisfied. What I had been asking also concerned why tasks were defined as they were, why particular standards appeared natural, and how the “I” using AI to understand and act came into being. These questions had not vanished from the text, but they had receded into the background of operational advice. In the writing, concrete had come to mean easy to fill in and easy to check. That interpretation organized the sections and the standards of self-review alike.

Adding more definitions at that point would have made the tables more complete, and the distortion more complete with them.

Revision therefore had to begin again with what the framework was trying to explain. The writing had treated the task as converting an existing theory into methods of use, while what the theory ought to explain still required inquiry. Two questions followed. If practice can also change how tasks are defined, theory must explain how that change happens. If concepts are to make the change intelligible, theoretical concreteness must lie in unfolding the relations: who adopts a determination in what situation, what follows from adopting it, and why it needs reconsideration. Practical guidance can then alter participation instead of merely adding operational items.

In this discussion, an interpretation first became the basis of work and then its object. It affected what we wrote, how we wrote it, and what counted as writing it well. Reconsidering it changed the writing that followed. The language in our hands helps a problem take shape; it can also fix that shape too soon.

“Make the object clear” is therefore insufficient. We must also understand how the object came to be articulated in this particular way.

Through Change a Way Opens; Through an Open Way Things Endure—The Yi Tradition, “Appended Statements,” Part II#

The Introduction to Hegel’s Phenomenology of Spirit questions a familiar arrangement: first examine whether cognition is a reliable instrument, then let it approach the object. Hegel asks what this already presupposes. Cognition stands on one side and what is true on the other, while the grounds for separating them remain unexplained. In the discussion of experience, consciousness confronts its object with the knowledge it holds and examines whether they agree. Where they do not, its knowledge must change; what it previously took to be the object’s essential determination may change as well. The standard used in examination thus enters the examination.⁷

To bring that question into this essay, consider a specific case. When we say an essay is “not comprehensive enough,” we are already treating it as something that should contain sufficient information and explanation. Missing details and unexplained levels readily become the next problems. That determination may be appropriate: technical documentation often needs its necessary conditions filled in. It may also conceal another demand. What experience must readers undergo to understand a relation that cannot yet be fully stated?

What comprehensive means cannot remain outside the work indefinitely. If a work is meant to present an unsettled relation between characters, and added explanation decides it for them in advance, the difficulty lies in the original standard: information has been supplied, but the unfolding relation has been cut short. We must reconsider what constitutes adequate expression in this work. What looked deficient may be precisely what its form needs to carry. In another essay that needs to state its evidence, the same omission could impede understanding. What the work is doing must ground the revision of its standard.

This revision passes through connected moments. The same passage remains before us, but the question changes from which information is missing to how narration allows a relation between characters to happen. With the question, the distinctions receiving attention change: perspective, order, pauses, and the timing of an explanation begin to guide revision. After deleting and rearranging, we must read the passage in context to judge whether the new arrangement preserves the relation of interest. Redetermining the object means distinguishing anew the relation we are working on; changing the essay itself requires actual revision. Thought gains force through this passage.

The same holds in natural science and engineering. A different model first changes how we describe and handle a system. Once the model enters design, computation, and execution, it may change how the system operates. Physical conditions act at every stage: whether a relation can be neglected, whether a computational result can be implemented, and whether a new behavior can persist must all find their answers there.

An anomaly may arise from an implementation error, a measurement problem, or an assumption in the model. Investigation, comparison, and argument tell us where a change is needed. The force of experience acquires concrete content through that work. At first, we see only that something has departed from expectation; later, we can identify which prior understanding has encountered what difficulty.

Hegel’s determinate negation supplies a point for understanding such change: the difficulty negates the earlier determination in a definite way, and a new understanding must form through that difficulty. In the concrete practices examined here, this calls for explaining why the object was grasped that way, where the difficulty arose, what remains valid, and what must change for understanding and action to continue. Reconstructing a problem must take up the grounds left by earlier knowledge and articulate relations it did not accommodate. Subsequent inquiry and action determine where that reconstruction leads.

An evaluation metric, for example, may capture one aspect of performance while omitting another important relation. New research retains the results supported by that metric, limits the scope of what it can establish, and develops criteria for the omitted relation. Negation therefore has two sides: explaining why the old metric cannot sustain the original overall judgment and allowing it to retain the narrower judgment it does support. The additional criteria must undergo their own examination.

The reasons for revising a standard thereby acquire content. Why does the original aim require a certain distinction? How did the existing standard erase it? How does the new arrangement restore it? Such a change responds to the original problem and makes it more fully understood. Changing standards merely to make an existing result look good deprives the new standard of this grounding.

The Preface to the Phenomenology also warns that familiar names—subjective, objective, subject, object—are easily treated as fixed points between which cognition moves without examining the points themselves. Its account of substance and subject, mediation and becoming, calls for the process of formation to enter our understanding.⁸ Human capacities, wishes, and ways of evaluating oneself also acquire their content in that process. How language and AI participate therefore belongs to our understanding of human beings themselves.

A stretch of practice may consequently involve changes in the determination of an object, standards of judgment, and a person’s relation to themselves. These changes have different rhythms and can affect one another. Someone initially devotes time to a certain kind of result; practice may lead them to reconsider what results merit that time. Understanding the longer course of research, creation, or a life requires room for this change.

Experience has weight because, after going through something, a person may no longer be able to understand it in the old way—or to understand themselves as they did before.

Do Not Let What Is Already Stored Harm What Is Yet to Be Received—Xunzi, “Dispelling Blindness”#

Research begins by entering into things. A report, a dataset, or an unexpected event can become material. Hesitation in a conversation or an unease in daily experience that resists description may also give a problem its first outline. Researchers need to understand concrete relations: what an event is connected to, where forces bear upon it, and which detail changes what matters.

Theory makes these encounters into objects we can recognize, compare, and continue to question. It supplies names and organizes relations. An operation that initially seemed isolated becomes connected to the properties of tools, the organization of labor, and other people’s needs; why it happens begins to appear. This is theory’s positive power to grasp reality.

Reality can also rewrite theory. If every detail is ultimately assigned to an existing category, theory has gained only more examples. Research requires material to change the grounds of classification and even relations between categories. Form and content can initially be discussed separately. Yet paraphrasing what a poem says, when its meaning happens in syllables and pauses, loses the experience created by a sustained sound or suspended meaning. This loss brings form into the constitution of content. To read the poem, theory must redraw its earlier boundary.

A theory thus has two connected modes of existence. It can be knowledge already achieved, helping us work with assurance. It can also reenter the making of thought and become an object requiring explanation, revision, and development. Learning moves between these modes: acquiring concepts and developing the capacity to let new material act upon them.

Researchers gradually learn to recognize what a theory makes visible and what it leaves unable to appear; which relations change when the same phenomenon enters another representation; why a seemingly irrelevant detail can require the problem to be rearranged. This capacity forms through repeated engagement with particular materials. The more precise a concept becomes, the more necessary it is to know what it has actually encountered.

The expression of theory brings this work to others. Explaining what enables something, what limits it, and whose needs it concerns lets people compare familiar approaches and may reveal another way to participate. Making theory accessible therefore has demanding work to do: let others see the relations, understand why a judgment holds, and continue thinking and choosing from there. An explanation succeeds through how it places the problem in readers’ hands.

For example, explaining stalled research as a lack of personal effort naturally suggests stricter scheduling. Further inquiry may reveal that the problem’s formulation combines two incompatible demands. No amount of time will then complete it as stated. Once that relation is explained, the researcher can separate the demands, redefine the problem, or prove that under particular conditions they cannot be jointly satisfied. Theory changes the understanding of the difficulty and the range of possible action.

At other times, there is initially no sufficiently clear object. Observations recur, but old names keep them apart. Through comparison and attempts, research identifies relations of mutual constraint or shared change and proposes a determination under which they can be studied together. Several phenomena separately treated as anomalies, for instance, may call for a joint inquiry into how information enters action. The new determination lets researchers select material, ask about causes, and design experiments distinguishing explanations. Scattered encounters gradually become an object for sustained inquiry. A new understanding of its essence must be explained through the key relations that make it this kind of object.

A framework must accommodate this creation. When a new determination explains previously scattered material and supports further inquiry and action, its theoretical value becomes concrete. Relations already worked through can become capacities, available to be unfolded again when needed. Further encounters enter and develop those capacities. Research accumulates knowledge and develops ways of forming it.

This work can also be read alongside Wheelwright Pian. In the “Way of Heaven” chapter of the Zhuangzi, Pian challenges Duke Huan, who is reading above him in the hall, from his own experience of shaping wheels below. He speaks of working “neither too slowly nor too quickly: I find it in my hands, and it answers in my heart,” and adds, “My mouth cannot put it into words, yet there is a measure within it.”⁹ That measure belongs first to the discrimination of the operation. Work at different speeds, and the wood and wheel respond differently. Even where the measure has not been fully articulated, its distinctions are exacting.

Pian’s difficulty reminds theory that the actual relations of a material may become understood only through the body and over time. Yet the words he does speak have already changed the problem of reading. Can the book in the hall preserve the skill practiced below? What must teaching transmit? Language has not delivered the whole craft, but it has brought a previously neglected relation into discussion.

More explanation and more experience can therefore support each other. Images, demonstrations, measurement, shared operations, and repeated practice can gradually make what was difficult to express discernible. AI may participate in this work. But if the entire exchange consists of rewriting the same explanation, the wood has still not entered, and judgments of craft still lack the thing they must answer to.

Letting reality change theory therefore requires arrangements through which material can actually enter inquiry: preserve details that do not fit expectations, hear those affected by an arrangement, and encounter a difficulty the concept has yet to explain. Then compare explanations, revise the corresponding conceptual relations, and use the new understanding in another attempt. Conceptual revision changes how inquiry proceeds, while the next attempt tests whether that revision has increased understanding.

Comprehend Its Changes, and Bring the Patterns of the World to Completion—The Yi Tradition, “Appended Statements,” Part I#

We can now state the preceding understanding more explicitly.

In this essay, practice means a process in which people use language, bodies, tools, and existing relations to understand and participate in things, sustaining, forming, or changing their relations to objects, norms of judgment and action, and their capacities and relations to themselves. Reading, research, making, interaction, and play can all be understood as practices. Sometimes they produce results, sustain a shared life, or let an experience not yet understood begin to appear. Judging a practice’s value and effects means asking what it allows to happen, how it affects participants, and whether those effects deserve to be sustained.

Objectification here means articulating differences so that they can be recognized, thought about, communicated, or handled. A research question brings certain quantities and conditions together; a diagram displays certain connections; a narrative gives a person a particular appearance. Each undertakes this work. Objectification has a positive and necessary role: shared inquiry and action would be difficult without some determination. Problems arise when it loses sight of its conditions, treats a partial grasp as the whole object, or fixes a changing person into a single property.

Norms are the demands within a practice concerning what counts as valid, fitting, well done, or worth doing. A norm may take the form of an explicit rule. It may also live in the measure of Pian’s hands, a reader’s grasp of a word, or expectations people have already undertaken in a relationship. Norms give differences their weight. Their origins, conditions of application, and justifying reasons can all become matters for discussion.

Together, these concepts yield five propositions. They explain how practice makes objects distinguishable, produces conditions for what follows, and changes its participants.

First, the determination of an object and the standards of its evaluation are related. When a passage is offered as a proof, the capacity of its reasons to support its conclusion is a constitutive requirement. When it enters a fictional character’s speech, tone, perspective, and relations join truth and falsity in the judgment of the work. The same words can acquire different statuses; their norms cannot be extracted from their wording alone. The actual properties of an object also constrain which determinations hold up, so choices always require reasons.

Second, practice can change some of the conditions that make it possible. An explanation is saved and adopted, becoming a premise for selecting later material; a promise is accepted by those involved, rearranging shared affairs; a discriminating capacity developed in practice makes harder work possible later. Adoption, retention, and implementation give products a role in organizing subsequent action. Time thus enters the interior of practice: available conditions have histories, and today’s ways of working may become tomorrow’s conditions.

Third, a transformation of practice can reach the relations between categories. Some changes require correcting a result or improving a method. Others leave the old relations among questions, representations, and evaluation inadequate to carry work forward. Reconstruction must account for the valid content of an earlier determination and its difficulties, and explain what understanding the new determination adds. Creation can actively bring about such change. A new work or a new arrangement for shared research may make relations appear that earlier classifications had no place for.

Fourth, the common structure of practices must be understood through their differences. Research, engineering, art, and interaction have their own histories, objects, and conditions of recognition. Where they overlap or transform, a demand may need another application: examining evidence in research grounds a knowledge claim; examining narrative perspective in fiction gives a character’s experience its place. A shared framework must explain these connections and distinctions and understand new forms of practice. It forms a whole when the relations illuminate one another while each practice can still unfold through its own object.

Fifth, self-formation occurs within relations of participation. People’s capacities, wishes, self-images, and positions of evaluation form and change through repeated practice. People can also claim, question, and alter some of these conditions. Understanding this formation requires following how participants acquire capacities, regard their achievements, seek recognition, and choose their commitments anew. The question of the subject develops from within these relations.

These five propositions belong to one framework because they can explain a course of change together. A representation makes an object jointly recognizable, and corresponding standards organize the work. Results are saved and reused, supplying the next attempt with new capacities and expectations. New material may reveal inadequacies in the original determination and standard. Through redefining problems, learning new approaches, and choosing where to continue investing effort, participants change their relation to the practice. One relation becomes a condition of another, allowing practice both to continue and to change.

The changes may intersect, or for a long time affect only limited parts of a practice. Established methods let work continue reliably; new problems make some of its relations objects of thought again. The framework helps us recognize what kind of change is occurring and put effort where it is needed. It can also help possibilities not yet formed acquire content.

When AI participates in this process, a new way of connecting language and action appears.

What Is Said Must Be Capable of Being Put into Practice—Analects, “Zilu”#

To tell an agent “work this problem through” is already to hand over a stretch of subsequent action.

Coding agents such as Codex can read files, modify code, run tools, and continue working in light of the results. Systems such as OpenClaw connect a model to a working directory, saved context, and tools. With the relevant capabilities enabled, an agent can also operate a browser: read pages, click, and enter text.¹⁰ In this arrangement, the model generates judgments and choices of action, the runtime connects those choices to tools, and the environment returns the results. What we commonly call an agent acquires its capacity for sustained work through these connected parts.

One request can encompass many actions that have not yet been individually decided. A person supplies the problem; the agent finds material, tries approaches, encounters errors, and determines the next step. Operations previously linked together by repeated human intervention can now unfold continuously within a task. The division of labor develops from receiving an answer on each occasion to jointly sustaining a stretch of work.

This change first touches the object. A directory handed to the system looks like a set of files; for ongoing research, the files also preserve connections between material and conclusions, earlier questions and new ones. Trustworthy material, established conclusions, and questions worth continuing have to be identified through these connections. Acting on an interpretation, an agent may turn it into groups of files, paths used by programs, and results of later retrieval. How an object is understood can thus enter how it is handled more quickly.

Imagine asking an agent to organize a research directory accumulated over many years. Filing by year can recover the chronology of the work. Reorganizing it around current questions can bring material from different periods into one inquiry. Moving all old drafts out of the way may sever the path by which a conclusion was reached. All three approaches can make a directory look tidy, yet they create different conditions for research. The determination carried by the word organize is progressively realized through moving files, updating links, and retrieving material later.

What is worth articulating in the collaboration is therefore which relations we want to carry forward. If the aim is to make a problem researchable again, work should be organized around the connections between material and judgment: preserve the path of key derivations, distinguish unresolved questions from established results, and enable later people and agents to reenter the problem. The beauty of the directory then comes from making the relations of research usable.

Norms also gain carriers through which they can act repeatedly. A statement of what counts as completion, supplied to an agent by its runtime, can guide the next action; a consistently used criterion can lead it to revise the same kind of product repeatedly. Task settings, checking programs, and renewed use of records bring standards into subsequent operations. Rules written for a system thus help organize future labor, and what a rule was meant to serve affects what the system keeps producing.

The scope of a rule matters concretely here. “Citations must be traceable” supports knowledge claims. Enforcing it as a demand that every judgment borrow the words of an authority makes an author’s own argument difficult to develop. “The program must pass its checks” verifies properties already defined. Treating that success as proof that the problem is solved leaves no place for relations absent from the checks. Good norms should help a system understand what kind of practice it is sustaining and bring newly encountered problems back into joint discussion.

The person initiating a task also gradually acquires a new position of participation. They can spend time on the formation of problems, comparison of explanations, and material that determines the next direction, while the system continuously performs operations ready to be handed over. This division enables one person to sustain a longer research process. It also gives personal participation a new meaning: redefining a problem, entering a decisive attempt, or allowing previously excluded material to change the work.

Delegation can extend agency. Someone who previously had only an idea gains conditions for trying, comparing, and revising through sustained execution. Each delegation also calls for an arrangement governing how results are adopted: on whose behalf the system works, who undertakes resulting commitments, and how those affected can object or change the arrangement. The more action can continue, the more these relations must be realized in shared understanding and actual permissions. Otherwise, work may continue without its initiator and those affected being able to deal with the situation it changes.

Computer environments are already part of real life. Changing a shared document affects a colleague’s work; publishing text enters other people’s reading; running a program consumes equipment and time. A file can be restored to an earlier version, while the reading and labor that have already occurred follow their own histories. An agent’s actions must be understood along these actual connections. One step in an interface can enter a long chain of social relations.

Time also changes. In a task already started, with usable tools and a means of continuing execution, work can progress while a person’s attention is elsewhere. On returning, they may face an altered environment: new material has arrived, a route has been tried, and the conditions of the next choice have changed. What they need to grasp includes what to let continue, where to reenter, and how unforeseen results can inform the next judgment.

An answer has now acquired the duration of action. Writing carries a person’s request beyond the present utterance, allowing it to keep happening in files, tools, and other people’s work. Longer delegation extends human power and makes its direction an object for practice itself to investigate.

Draw on the Body Nearby and on Things Farther Away—The Yi Tradition, “Appended Statements,” Part II#

This course of action is also beginning to connect with machine perception and movement. Gemini Robotics 1.5, announced in 2025, combines a model responsible for planning and tool use with one that converts visual information and instructions into robot actions.¹¹ Such research systems let us ask further: if these capacities reach wider everyday use, how will language enter movement, contact, and cooperation in shared space? The household scenarios below explore new conditions of cooperation and the relations they may change.

At that point, “tidy this up” will act upon tables, chairs, clothes, and the paths people take through a room. How far this extends, what counts as tidy, and which objects should stay where they are will acquire content in the operation. Ambiguity can remain suspended for a while in a discussion. Once an arm begins to pick things up and wheels begin to move, one interpretation has already acquired spatial force.

Operations in a computer can already have real consequences. Robots bring another set of relations close to the action. Objects have weight, friction, and fragility. People move, tire, and suddenly reach out. The machine’s perceptual range, speed, and ways of making contact determine what cooperation is possible. An agent enters the world through this particular body, which helps determine how it can understand and carry out a task.

Pian’s “measure within” finds another echo here. Language brings a task to hand; the material’s response gives the movement its measure. For robots with the relevant sensing and control capacities, sensor information participates in adjusting movement, and the task is completed through continuous perception and action. Whether an account of a movement is adequate depends on how it joins this process.

Consider a future household. A robot brings frequently used objects to a person with limited mobility, making it easier to begin reading, making something, or receiving friends. Less effort is spent fetching and carrying. Activities that once required someone else to be constantly available may become activities the person can initiate themselves. New conditions of cooperation extend agency, giving “what I can do” new practical content.

Whose body determines ease of use remains a question within this benefit. A cleared tabletop looks orderly; putting a cup on a high shelf may leave its regular user unable to reach it. If tidying is meant to support their daily activities, the position of things must be considered in relation to their body and movements. The object of tidying expands from the arrangement of items alone to relations among things, bodies, and everyday action. These relations determine whether a particular order deserves to be maintained.

Shared living develops the object further. One person’s instruction may move something another person is using. A saved preference may become an order repeatedly enacted across the whole room. With a machine occupying a position from which it can act in space, the person giving instructions acquires power to arrange that space through it. Who can make requests, whose habits of use become visible, and who can change an arrangement already in force help determine how this power enters shared life.

Norms become part of the cooperation itself. Keeping ordinary objects within reach of those who need them, allowing affected people to understand and adjust changes in shared space, and coordinating machine movements with human movements provide conditions for living together. Stable ways of working let people act with confidence. When circumstances change, participants need ways to introduce new demands. Discussion and trial can bring those needs into the arrangement and allow another stability to form.

Embodiment also changes the rhythm of interaction. A response on a screen is usually taken up in reading; an action in a room enters waiting, making way, handing over, looking together, and doing things together. A glance or gesture may revise the original request. An object slipping may force a movement to change. Discernible forms extend from sentences to sounds, postures, and motion, and mutual understanding gains its concrete measure in these exchanges.

If this cooperation gradually enters labor, tasks themselves will differentiate anew. Skilled workers can turn part of their experience into arrangements a machine can execute, while continuing to recognize materials, handle anomalies, and develop new ways of working. Some people previously excluded by physical demands, distance, or environmental conditions may gain new opportunities to participate. Institutional distributions of equipment, time, and results will help determine who receives those opportunities. Robots entering labor change the organization of capacities and the conditions of participation.

These changes also enter a person’s relation to themselves. Someone who works over time with a machine that can answer and act may incorporate it into “my way of working,” even into the immediate experience of “I can.” The assistance it provides, the trust placed in it, and the abilities the person wishes to keep learning acquire different weights through daily use. The human body has not performed every movement on its own, yet the person’s life has changed through the new cooperation.

As agents move from computers into physical space, the theoretical object of practice develops with them: how language organizes action, how action changes an environment, and how that environment supports or rearranges people’s capacities and relations. In everyday use these processes gradually intertwine, becoming a question of how humans and AI live together. The person who says “I” encounters themselves anew within these changes.

People Can Enlarge the Way; the Way Does Not Enlarge People—Analects, “Duke Ling of Wei”#

A text whose words remain unchanged can change a person’s question when reread. Teachers, friends, and collaborators also alter understanding through their responses. Human capacities have always grown within such relations. The sustained collaboration enabled by agents connects generation, recording, execution, and evaluation more closely, bringing the capacities and standards formed there into daily life more frequently.

A suggestion about “good writing,” once saved, may govern later revisions. A definition of a research problem, once adopted, may determine what material is selected. A system that regularly acts on someone’s behalf may help change what they are willing to begin next. The wider their adoption, the more closely these formulations’ power to shape subsequent work deserves examination.

This lets us distinguish two kinds of acceleration. Completing familiar work faster is chiefly an efficiency improvement. At other times, saved time and sustained responses make previously unsustainable work possible: material accumulates, attempts recur, and a question still forming can develop from earlier results. If the work gradually yields reusable arrangements of material, questions that can be pursued, and ways of participating, its form has changed as well. Qualitative change should be recognized in that new organization, rather than inferred merely from faster responses or more products.

Such changes are exciting. What was once only a wish can become a work. Research once contemplated from a distance becomes something one can enter. With the relevant systems, materials, and tools, people can develop reliable shared capacities. These capacities have conditions, just as literacy depends on linguistic traditions and experimental ability depends on instruments and other people’s work. Explaining those dependencies helps us understand how to develop the capacities further.

The same achievement also enters a person’s relation to themselves. Completing an essay can become evidence that “I can write an essay like this.” Claiming it in this way affects the next expectation. An image emerges: someone thorough in thought, swift in action, able to grasp many problems at once. A person may admire this image and begin to demand that they match it at all times.

In his account of the mirror stage, Lacan connects identification to the transformation that occurs when a subject assumes an image. The ego’s unity is organized through an external image, with a tension between its apparent wholeness and the actual process of formation.¹² Bringing this question into AI practice means examining how an achievement makes an “I” available to be claimed. This mirroring relation has a constitutive role, extending far beyond a screen faithfully reflecting an already existing self.

The system can also evaluate achievements and describe a user’s style, strengths, and weaknesses. A person may acquire both an image and a standard for viewing it. Lacan’s distinction between the ideal ego and the ego ideal lets us examine them separately: the ideal ego concerns an apparently whole image available for identification; the ego ideal concerns the symbolic position from which that image merits affirmation—the gaze and demands under which I hope to be recognized.¹³ The sense that I am capable and the question of which capability deserves affirmation thus have different grounds, even when they arise from the same achievement.

What sort of person I want to become and the position from which I hope to be regarded as valuable may lie close together, yet they are different questions. The recognition a researcher seeks may come through standards of argument and contribution intelligible to peers. The readers a writer anticipates also help determine what counts as mature expression. If AI continually invokes a certain standard in collaboration, it may participate in transmitting and rearranging these relations of recognition.

Consider thoroughness. It can help research supply necessary conditions and make judgments more reliable. As it becomes a standard for every expression, however, a feeling still taking shape may be treated as deficient. Every hesitation must then receive a full explanation; every passage must answer all possible objections in advance. Someone can become increasingly capable in appearance while finding it increasingly difficult to tolerate not yet knowing their own mind.

An ideal image can also attract someone toward learning. Better work encourages further reading and practice; as capacities grow, the person can judge more accurately what they previously achieved with a system’s help. These developments depend on how the result is subsequently used. Is it an image one must constantly match, or a starting point that makes another attempt possible? Following actual adoption and practice lets us distinguish how capacity, identification, and evaluation interact.

The boundaries of a person must therefore be understood separately through capacity, attribution, and identification. The ability to complete work may depend on systems and tools; results and responsibility acquire attribution through shared arrangements; identifying something as my thought also passes through reading, judgment, and actual adoption. A robot may become part of someone’s capacity to act, while generated words remain something they cannot claim as theirs. These boundaries may coincide or diverge. Concrete cooperation, recognition, and disagreement show which connections have formed.

In “Science and Truth,” Lacan relates the division of the subject to the unconscious and to the relation between knowledge and truth. The question of the subject thereby exceeds a shortage of information about the person.¹⁴ Someone’s knowledge of themselves, including knowledge organized with AI, cannot exhaust the subject who speaks, identifies, desires, and acts. A more accurate profile has its uses; it does not make the subject an object whose description can finally be completed.

This essay draws on Hegel’s account of mediation and becoming to understand how capacities and relations to oneself form through practice, and on Lacan’s divided subject to understand why a person cannot wholly coincide with an established self-image. Each addresses a distinct question here. A capacity one can claim gives a relatively stable answer to what I can do; the person claiming it may still encounter, in desire, speech, and action, something that answer cannot accommodate. Formation gives a person provisional determinations, while noncoincidence allows them to become questions again.

How this noncoincidence enters change must also be followed through particular events. New material undermines an earlier judgment; another person’s response reveals a partial self-account; a personal attempt loosens the claim that I can only do this. Through such encounters, someone can reconsider their commitments, try another arrangement, and examine its results. The language and powers used to begin the change also come from existing relations. Agency appears in this work, with its histories and resistance, as people participate in changing some of the conditions that constitute them.

Recognition brings human formation into another relation. In his account of self-consciousness, Hegel places recognition between self-consciousnesses. His analysis of formative activity also brings relations to objects into self-understanding.¹⁵ An agent’s evaluations may transmit a community’s standards. The recognition a person seeks also concerns who responds from their own position and what that response commits each party to. Instrumental responses, the transmission of evaluations, and mutual recognition each involve relations that need to be examined.

Especially in shared life, others respond from their own circumstances. An AI-generated analysis may help us understand another person’s views, but cannot consent on their behalf. A highly rated achievement may still lack recognition from the relevant community. The different sources of responses keep self-formation connected to the concrete world.

When one window can explain, evaluate, and execute in succession, it may feel as though the order of language itself is answering. A reply is taken as what everyone would judge, or advice as what a mature person should do. The system is then trusted as a representative of shared standards as well as a source of information. This brings in the question of the big Other: which languages, rules, and conditions of recognition make words count, and whom people take to answer on their behalf. A system gains authority through this trust. Its transmitted norms, supporting institutions and technologies, and the grounds on which people grant it authority to evaluate all require concrete inquiry. The answer in the window connects with a wider linguistic and social order.

AI is thus a power available within practice and a condition practice needs to understand. Its positive possibilities cannot be measured solely by how much work gets done. They include which questions can be sustained, which capacities can grow, which experiences finally acquire expressive form, and which relations between people can be rearranged.

The Gentleman Harmonizes Without Being the Same—Analects, “Zilu”#

Returning to whether language should follow AI’s norms, we can now ask more precisely: how does a form of expression help this practice develop fully? Under what conditions does it change what the practice originally sought to address?

Norms first have force because they make a practice possible. A proof needs reasons that support its conclusion. A design needs components that work together under the relevant conditions. A shared arrangement needs participants able to understand and undertake it. Without these demands, research, engineering, and cooperation lose their grounds for continuing.

There are also different moments within one practice. Research needs unverified conjectures because existing knowledge has not answered new questions. When conjectures become conclusions, they must meet corresponding demands for proof and evidence. Requiring everything to be complete at the outset makes new questions difficult to begin. Retaining the freedom of conjecture indefinitely without undertaking examination makes knowledge difficult to achieve. Norms should allow these two kinds of work to support each other.

Rigor in form also bears upon different things. A sentence in a novel can be simple in factual terms and finely discriminating in its relations. The difference between “He came back” and “He finally came back” exceeds the addition of a word. Finally brings waiting and expectation into the narration. Who has been waiting, and who regards this return as warranting that word, must find a position within the work. Writing hesitation accurately may require deleting an apparently thorough explanation.

Literary discrimination and scientific rigor can both be understood as answering to their objects, while the concrete ways of doing so remain distinct. A work needs the experience its form creates to become discernible. A scientific claim needs argument and material that support its scope. Factual judgments within a work still answer to factual demands; expression within research also affects how a question becomes visible. Overlapping domains retain their differences.

Interaction puts the word object to a further test. Facing another person, we can study their situation while also relating to them. As a participant, they have needs, judgments, and the possibility of refusal. Their response can change both the object under discussion and how discussion should continue. If a supposed consultation has predetermined its result and leaves only the work of securing acceptance, it gives no place to the response it claims to welcome. Clearer language cannot repair this alone; the arrangement of consultation must change.

Such criticism begins within the practice: how does it actually treat what it claims to be doing? Internal consistency, however, cannot supply all its legitimacy. An arrangement that excludes certain people can be perfectly orderly. The clarity with which “Rectifying Names” distinguishes noble and base does not establish for us today that those ranks should exist. The purposes of a practice, eligibility to participate, and distribution of benefits must also be open to questioning.

The grounds of norms therefore include several related kinds of reason. An object’s properties and conditions constrain what can work. The identity and commitments a practice undertakes determine what counts as doing that thing. Value judgments concerning the lives, relations, and consequences worth pursuing let us judge whether it should continue. Historical provenance explains how a standard gained standing; justification explains why it merits adoption. Neither inquiry can replace the other.

When demands conflict, we must return to the relations among these reasons. Asking how to determine the object means asking which understanding best explains what we are doing and which distinctions cannot be omitted. Physical constraints, the situations of those affected, and experience established through a work’s form all supply grounds. Even after recognizing the same situation, people may disagree about whose needs deserve priority or which costs to accept. Reasons for the value judgment must then be stated, with those bearing the consequences able to enter discussion. Redetermining an object can clarify conflict without deciding every choice in advance.

For example, a team can make a limited attempt before all the evidence is available in order to explore a research direction promptly. Reasons to act come from existing grounds, the cost of the attempt, and what can still be learned; the knowledge claim remains unsettled. Efficiency and honesty can hold together here. If the attempt places a new practical burden on someone else, their participation also belongs to a legitimate arrangement.

Accuracy, creation, efficiency, interaction, and revisability often support one another. Where they cannot all be achieved, concrete objects and circumstances give reasons for the choice. Once agents enter sustained action, such choices also enter rules, interfaces, and habits of operation. Discussion of norms should therefore follow them into how they organize things: what do they make easier to do, and what becomes difficult to bring in?

The earlier sevenfold classification can still help us recognize these differences. Its types are points of entry, not official categories into which lives must be registered. A piece of writing can conduct research, create a work, and develop self-understanding at once. An engineering discussion may change relations among its participants. When a new form appears, the framework should investigate why it is organized that way rather than begin by assigning it a box.

Establishing norms for language thus involves two connected kinds of work: giving valuable relations a reliable expression and making determinations that no longer fit available for questioning. The first lets people continue with confidence. The second lets that continuation depart from the old path.

What One Learns on Paper Proves Shallow; to Know It Fully, One Must Put It into Practice—Lu You, “Reading on a Winter Night, Shown to Ziyu”#

A sustained hypothetical example can show what this understanding changes in actual research.

Imagine a researcher improving a vehicle platoon’s following control. Working with AI, they check derivations, run simulations, and adjust parameters, evaluating performance by the mean squared following error over the full run. After one round, the new parameters lower this value, and the researcher prepares to report “improved performance.” A closer look at a braking episode reveals a local response worse than expected. Better average performance across the run has not removed the transient problem that actually concerns them.

The reduction in mean squared error establishes an improvement in the performance that this metric expresses. Judging whether it adequately represents the platoon behavior of interest requires reopening distinctions the aggregate statistic leaves out. Local peaks, minimum spacing, and the duration of anomalous behavior may carry different demands. The earlier result thereby acquires a more accurate place.

The next work is concrete: return to the relevant trajectories, identify the observed quantity, check whether implementation agrees with the model, and establish whether the anomaly occurs within allowed conditions. If measurement or implementation is at fault, correct that part first. If the anomaly genuinely occurs under the specified conditions, ask what the metric and model can each explain. These checks turn an initially vague disappointment into a located research difficulty. They mediate how experience changes judgment.

This investigation can lead to different tasks. If the model already represents the relevant process adequately, the difficulty may lie in an optimization objective that omits local requirements. The objective and constraints then need investigation. If separate material indicates that the crucial phenomenon concerns the timing and use of message updates that the model does not represent, while communication effects were reduced to a constant delay, the model itself needs reconsideration. An inadequate metric and an inadequate model are distinct problems; the first does not establish the second.

Following the second possibility, work initially centered on a control law may now need to investigate relations among measurement, message handling, estimation, and execution. The next step can compare representations, identify which timing information must be retained, and propose experiments that distinguish explanations.

The research object has changed: a question initially centered on a control law and its effects now includes when information becomes available, how it is handled, and when it enters execution. Added structure deserves consideration because the material calls for such an explanation; subsequent analysis and experiments must determine whether it resolves the difficulty. Representations already capable of answering the question remain in use. The relations required by inquiry determine how simple or complex its model should be.

AI can undertake much of this work: reading code and model descriptions, organizing explanations, helping with derivations, running programs, and comparing results. The researcher needs to participate in relations that determine the inquiry’s direction: what do we want to know, what does a representation omit, and which result could change the present judgment? Participation can take the form of shared reasoning, trying a crucial part personally, rereading material, and examining key transitions. Reliably completed operations need not all be repeated by hand.

A deeper step is to give a problem not yet formed the status of a research object. At first there is only a sense that “this response is wrong.” Later it is defined as a relation between information and control under particular conditions, making hypotheses, models, and experiments possible. The new object is made through relations among existing questions, concrete materials, and new representations; it is not selected from a dictionary. Whether it deserves to be retained depends on its capacity to support fuller explanation and further work.

Standards of evaluation change as well. A choice previously made by one metric now has to be judged through the new relations of the object. We should explain which requirements can hold together, which involve costs, and why the present tradeoff is warranted. A standard cannot be tailored to a solution simply because it has already been built, or exempted from discussion simply because it was used before.

This process may also change how the researcher claims their work. Initially they want to be someone who can improve an existing metric. Later they may begin to value work on the determination of problems, modeling relations, and the scope of evidence. How they allocate time and assess later achievements gradually reveals the change. Their expectations of themselves have always helped determine where they are willing to spend effort.

Research can also begin with a question that was previously difficult to sustain. Suppose the same researcher wants to compare several control schemes but lacks the time to organize their different assumptions, implementations, and results enough to make comparison possible. With AI participating in organization and implementation, the work gradually becomes repeatable. The researcher begins to find that the worthwhile question extends beyond which scheme gets a higher score. It also concerns the observation boundaries and environmental conditions under which their behaviors can be compared, and when one scheme can replace another.

A new research arrangement is needed: specify which behaviors and conditions the comparison retains, analyze the schemes under those conditions, and explain which earlier judgments a substitution would change. Similar error values, for example, do not by themselves establish interchangeability under different communication conditions. The comparison must specify relevant inputs, observed quantities, and environmental conditions together. For this researcher, improving one controller develops into studying a set of conditionally comparable schemes. New questions, representations, and sustained work gradually give this object its form.

The researcher also undertakes different responsibilities: understanding how each scheme acquires meaning, judging why the comparison holds, and enabling others to continue the work. Shared capacity extends the range of research, which in turn changes the individual capacities that need to develop. Organizing different results into a justified comparison becomes part of what that capacity must accomplish.

Research guidance consequently has several depths. Familiar problems may need only established methods applied directly. When material presents a difficulty, different explanations should be made distinguishable. If the difficulty reaches the determination of the object, representations and questions need reconstruction. New capacities can also make a previously unsustainable direction researchable. The actual problem determines the depth of reflection, which must take effect in corresponding work.

A new determination must then enter models, materials, arguments, and experiments. A different representation of timing calls for corresponding changes in programs and observations. A new hypothesis calls for work capable of distinguishing it. An agent can pursue these operations continuously, while the researcher works with it to maintain the relation between the new problem and each step. In this continuity, theory becomes a real condition of research.

Earlier results are also involved. Existing code, figures, and conclusions may remain useful while their purposes and valid scope need adjustment. Which results to preserve, which dependencies no longer hold, and how new observations enter subsequent questions all need to be worked out. Revision then has intellectual content and a basis from which work can continue.

Finite experiments describe performance under the conditions tried. Mathematical proof establishes relations under given assumptions. Implementation and deployment connect those relations to concrete computation and operating environments. The formation of a problem and the formation of evidence thus enter the inquiry together. We need to know why a method is used at this point and which judgment it establishes. A dedicated treatment of these technical relations appears in “From Models to Engineering Judgment”.

For everyday collaboration with AI, this means making the formation and reconstruction of problems part of the shared work. Instructions need not supply every answer in advance. They can tell the AI which relation is under joint investigation, which judgments remain provisional, and what material should prompt reconsideration of the original determination. Such a handover grows from an understanding of practice. Its wording can be brief without the thought it carries becoming shallow.

Writing Does Not Exhaust Speech; Speech Does Not Exhaust Meaning—The Yi Tradition, “Appended Statements,” Part I#

Writing makes another kind of change easier to see.

Imagine an author writing a reunion between two people. In the draft, one gives no explanation for leaving; the other asks no questions and simply pushes a glass of water across the table. To clarify their relationship, AI adds the reasons for the departure, the psychology of the silence, and the meaning of the gesture. Readers no longer have to wonder, but a relation not yet settled has been explained into a conclusion.

Calling the problem “too much AI style” is insufficient. That diagnosis readily leads to deleting certain words, shortening sentences, or adding colloquial language. What needs examination is how explanation has been arranged in relation to the experience of the work. If the characters’ relation is happening in that silence and gesture, a narrator who speaks it all for them in advance changes the work’s object.

The author can compare versions with AI, keeping the events and changing perspective, order, and degree of explanation. Then read in context: what do readers know before the water is offered, which interpretations remain possible afterward, and has the narrator already supplied forgiveness or refusal that has yet to occur between the characters? Such comparisons make discernible both the relations omission preserves and the judgments explanation adds. The author develops a capacity to recognize what different ways of writing bring about.

The new understanding can then enter later writing. The author can ask AI to check facts, distinguish word meanings, organize material, and propose narrative arrangements, judging them by what the work is doing. Omission may allow a relation between characters to appear; full explanation may make an argument intelligible. Each finds its measure in the content being written.

The rewriting of this essay follows the same relation. If it is to explain how practices form and change, its concepts need to acquire content through the unfolding of those relations: how an object appears, how a standard acts, and how people act within the process. Writing here undertakes the work of bringing relations into intelligible form.

Learning makes the position of participation a more explicit question. A good explanation can bring someone close to material while quickly giving them the feeling that they have understood. To acquire a command of reasoning, their own attempts need to encounter its key relations: explain why a step holds, compare a similar case that fails, or change the conditions and see how much of the original claim survives. AI can organize these attempts, give feedback, and help locate the difficulty.

The capacity being formed relates to the product, yet the product cannot represent it on its own. Completing a proof together and being able to explain its decisive step may support one another as goals. The first requires checking the conclusion and argument; the second also requires chances for the learner to attempt, err, and explain. If the immediate need is reliable use of a tool, learning every internal detail may be inappropriate. If a relation determines later judgment, it deserves participation. The practice one expects to undertake determines where capacity needs to form.

Learning also means allowing an object to change an existing explanation. When reading a philosophical text, one can first use AI for an introduction, then return to the text with that introduction in hand: which sentences has it left unexplained, and which turns cannot fit its framework? The guide makes reading possible; the original makes the guide questionable. Through this movement, a reader may develop new questions instead of accumulating ever smoother paraphrases.

Changes in everyday life often look less like research, yet they too concern the determination of objects.

Consider another shared arrangement. Several people want more time together and ask AI to help organize their schedules, filling free periods efficiently. Total time together increases, fulfilling part of the original aim. Gradually, however, being together starts to feel like rushing from one appointment to another. If the difficulty is still understood mainly as insufficient use of time, the next step will be further schedule optimization.

Another understanding would ask what shared life the arrangement is meant to form. Perhaps people need cooking, walks, an occasional conversation while each does their own thing, and time spent together with no topic planned. These bear a different relation to time from consecutive blocks of activities. Unfilled time is not automatically waste, and participants are more than resources whose free hours are handed over to a plan.

Some prefer explicit arrangements; others need room to be alone. Work and care impose constraints that cannot be moved at will. Letting these situations actually enter the discussion makes it possible to redefine the shared object: what kind of being together do we want to form, and how can different needs have a place? The response each person brings from their own life can give the earlier arrangement new content.

The new understanding can change what is done concretely. Schedule regular activities and keep time without a planned agenda; let shared work such as cooking become part of being together rather than always treating it as a burden to clear first; let participants introduce needs that differ from the original plan. Afterward, hear how each person experiences it, whether reserved time is genuinely available, and whether labor falls back on a few people. Such feedback makes more time together concern how companionship happens and who sustains it as well as the number of hours. Organization and evaluation change with this shared aim.

A person’s demands upon themselves may also be touched. Someone who has always understood a worthy self as never wasting time and always producing results may find a looser arrangement unsettling. That unease can be described and understood through the situation itself. What actually happened in this time? Who gained what kind of companionship? Am I willing to let such things continue? Judgment can then come closer to the life being lived.

Self-formation here includes a change in a relation of evaluation. People gradually learn which time has value, what they are willing to undertake, and how to live with others. The change acquires content through living. Disagreements and unresolved matters remain for later life together to answer.

AI can help organize constraints, compare arrangements, and preserve an understanding formed together. If it also executes schedules or moves objects, that understanding must enter operating rules: reserve time for shared presence, place things to support different users, and reconsider arrangements when circumstances change. Saved agreement also needs a history showing what those involved have accepted and what remains open. Connecting discussion, execution, and renewed discussion in this way can make a new way of living sustainable.

These examples suggest changes in how we participate day to day. Let problems still taking shape enter collaboration, and let different representations be tested by their objects. Arrange personal attempts where a capacity is meant to develop, so that AI’s work becomes a condition for further learning. Allow new experience and other people’s responses to affect existing arrangements, keeping records, goals, and self-expectations grounded in particular situations.

These ways of participating can give people greater powers and more uses for them. In reading a poem, playing a game, or talking with someone, what is sought may lie in the experience itself. Understanding how such experiences make life worth living can help us arrange time, tools, and our participation with one another more concretely, allowing these things to happen fully.

Illuminate What Has Gone Before, and Examine What Is to Come—The Yi Tradition, “Appended Statements,” Part II#

When a stretch of practice ends, what it leaves behind continues to act.

A research account leads later people along a particular question. A habit of expression makes some experiences easier to articulate. A shared arrangement gives life together a relatively stable rhythm. Achieving such results takes time, and being able to use them later is an important capacity. People cannot begin from nothing every time.

Yet the conditions left behind have histories. A rule may have been established for a particular engineering judgment; a way of writing may suit only one essay. An account of oneself may come from a life that has already changed. When these enter new practices and their origins and scope are forgotten, past validity can readily become present obviousness.

What deserves preservation is a record that lets someone later reenter the problem: what was adopted, what it sought to explain, why it was defined that way, and what remains unresolved. Agents using that record again need to distinguish its scope. A demand to explain evidence fully may continue to help argumentative writing without fitting a fictional character’s silence. Keeping key reasons and relations grounds reuse and helps revision locate what needs to change.

Revision then enters what follows. A changed research object calls for corresponding changes in models and experiments. A changed shared arrangement must let participants actually live through the new relations. A different understanding of self-evaluation must gradually give daily choices corresponding grounds. Earlier commitments and consequences already incurred still need to be taken up. At these points, changed thought encounters the test of reality.

The framework itself participates in this process. Relations among objects, norms, and subjects need concrete content in a particular inquiry, piece of writing, or shared life. Concepts help us enter these things; the things change the concepts. Understanding can accumulate while remaining open to problems yet to appear.

The framework’s coherence as a whole rests on these connections. Delegation on a screen, cooperation in a room, and people’s claims upon capacities and lives all involve how a determination becomes a condition through action and changes through new responses. This shared question lets different practices illuminate one another; their particular technical, bodily, and social conditions require separate inquiry. We preserve achieved understanding and develop it through new participation.

Return to the story of writing’s invention. Grain falls from heaven; ghosts weep in the night. Writing makes human ability and the ordering of the world a question together. Liu Xie moves from the patterns of heaven and earth to human cultivation. The transmission of talismans and diagrams brings those who acquire their power into a new life. Pian brings the measure of his hands before a book. Writing can reveal, determine, and transmit; it also encounters what has yet to be fully said. Its power lies in these concrete relations.

Today’s answers come from different places: a system supplies a reply, a tool performs an operation, people we live with express their needs, and actual results support or obstruct expectations. A system can reply quickly; things still unfold under their own conditions. Following these different responses tells us how an understanding took effect, where it needs to change, and what a newly acquired capacity makes possible.

The person who says “I” is in this process. Beginning with language already there, they give it new meanings through practice. Formed in relations, they participate in changing relations. The worth of human–AI practice can gradually be answered by a life of this kind: understanding the objects we face more fully, becoming more capable of worthwhile action, and better apprehending the relations we are forming with others and with the world.

Grain rains from heaven; ghosts weep at night. We have already received writing. How we answer, and how we allow others’ answers to enter what we do, will also help shape who we become.

Cover: an original conceptual image generated with AI for this essay, connecting an ink line on paper with its echo on a screen.