AI holds the answersdo you know the questions

A question is usually treated as a way of asking. It is better understood as the moment consciousness arranges itself around a gap, because attention turns toward what the question opens. A vague question opens a vague place, and the mind fills it with the most familiar answer it holds, which is why so many of our own questions return the same thoughts.

The first hour takes the question apart. A question has four parts that can be stated deliberately: what is wanted, in what form, under which conditions, and against what standard. The standard is the part most often left out and the one that decides everything, because a question without a standard admits every answer. We look at what a leading question does to the one who asks it, and why a question held for a while returns more than one answered at once.

The second hour is your own question. You bring one question you keep asking yourself — about your work, your learning or a decision — and it is taken apart here and rebuilt: the wish separated from the inquiry, the premise tested, the standard written down before the thinking starts. Then it is held, the answers are sorted into what is true, what is false and what remains unknown, and what gets revised is the question rather than the answer.

You leave with the habit of composing the question, and with one question of your own that opens a place worth entering.

A prompt is usually treated as a way of asking. It is better understood as a description of the answer you are willing to accept, written before that answer exists. Most of what people call a bad output is an unstructured request: the question carried no shape, so the machine supplied one of its own — the most frequent shape it had met — which is why so many answers arrive competent, general and useless.

The first hour takes the ask apart. A question has four parts that can be stated deliberately: what is wanted, in what form, under which conditions, and against what standard. The standard is the part most often left out and the one that decides everything, because a request without a stated standard cannot fail, and what cannot fail cannot succeed either. We look at what a leading question does to an answer, why an example teaches form faster than any instruction, and why the same request phrased twice returns two different things.

The second hour is your own material. You bring one instruction you have already given a machine — a summary, a piece of research, a draft, an email you send weekly — and it is taken apart here and rebuilt: the wish separated from the rule, the right altitude found between them, the standard written down before the work runs, the examples chosen for the form they carry rather than for what they say. Then it runs, it is judged against the standard you set, and what gets revised is the question rather than the answer.

You leave with the habit of composing the ask, and with one instruction of your own that works and can be shown to work. Most people, after three hours of this, stop rewriting outputs and start rewriting questions, which is the whole distance between using a machine and working with one.

Seeing is usually treated as recording. It is better understood as a selection, because attention weights what competes for it and the rest is attenuated before it reaches awareness. What is in view is not thereby seen, and most of what we call a good eye is a trained attention.

The first hour is about the gate. We look at why a fully visible change goes unnoticed, why a still eye stops seeing, and why training changes what gets through rather than what the senses detect. Naming is the work here, because a scene you cannot describe is a scene you did not fully perceive.

The second hour observes. One scene — an image, a room, a sound — is looked at in turns: first freely, then with attention placed on one feature and returned each time it wanders, then described in full and compared with the scene itself. What was missed is recorded, and the observation is repeated until the description holds.

You leave with one scene perceived in full and the routine that perceived it, which serves every scene after it.

A page is not a stack of sections. It is a set of relations — what depends on what, what is read first, what may be removed without the rest collapsing — and a machine asked for a page without those relations returns fragments, because fragments are what the request described.

The first hour is spent reading. Three pages you admire are opened and taken apart until you can name what is actually happening inside them: where the eye is sent and by what, which decisions repeat, what each of them refuses to do. Naming is the work here. A reference you cannot describe is a reference you cannot use, and most design instructions fail because they carry an impression rather than a structure.

The second hour builds. The elements are gathered first and then reduced — what the page uses, and what it deliberately is not, stated before anything is made. The page is then written as one description, complete enough to produce it and open enough to surprise you, and several whole versions are generated from it rather than one attempt polished repeatedly. Whole versions can be judged against each other; a single attempt can only be tinkered with. The one taken to a finish is the one whose winning axis can be named, because that reason is the method and the only part that transfers to the next page.

You leave with a finished page and the description that produced it, which is the thing worth keeping. The page serves one project; the description serves every page after it.

An imagined image is usually approached as something that simply appears. It can be built deliberately, part by part and relation by relation, and once built it can be held, turned and moved in the same way the eye explores a real scene.

The first hour is construction. The image is described in words: what it contains, where each part stands, what stays fixed and what changes. Its beginning and its end are held as two complete pictures, because a movement between two unfinished pictures cannot be followed. We look at the two frames of space the mind uses — one around the body, one among the things — and at why time is so often pictured as distance.

The second hour holds and moves. One image is carried from its description to a stable picture, then moved, then described again from memory and compared with the first description. What was measured and what was only felt are separated, and the image is refined one feature at a time.

You leave with one image you can build again and the sequence that built it — a way of imagining that can be described, taught and repeated.

Motion is usually approached from the wrong end — an element is nudged until it feels right, and what was learned cannot be repeated because it was never stated. A movement can be conceived in words before a single value is set, and everything after that becomes judgeable.

The first hour is conception. The whole movement is seen and then described: what changes between its beginning and its end, what stays fixed, what the change is for. Both halves are then designed as complete stills, because a transition between two unfinished things cannot be judged — you would be reading the transition and the design at once, and would not know which one is wrong. The description is written exactly enough to be built, with its openings left on purpose rather than by accident.

The second hour builds and judges. One element is carried from that description to a finished movement, in several whole variations rather than one adjusted repeatedly. Then the harder separation: what is measured against what is only recognised. Duration, distance and easing are measured; weight, intent and rightness are recognised. Most disagreements about motion are two people judging on different sides of that line, and naming which side you are on ends the argument in a sentence. The feeling is tuned last, one variable at a time, where the difference between an invisible change and a wrong one is learned.

You leave with a finished movement and the sequence that produced it — a way of describing motion that a machine can build and a person can argue with.

Everyone who learns for long enough arrives at the same problem: there is more than can be held, and what was understood clearly last month returns only in pieces. Memory is the structure that stands between those two, and most people build it by accident.

The first hour is about what memory is for. A representation has two strengths — how deeply it is stored and how easily it returns now — and performance follows the second. We look at why rereading feels like learning and teaches less, why spacing and testing outlast repetition, and why material grouped against what you already know is held in fewer units.

The second hour structures your own set. You bring material you are learning — notes, a subject, a language, a craft — and it is given a structure here: parts named by the questions they answer, relations drawn between them, and a schedule of retrieval set. The measure is stated before the work: a question you could not answer this morning, answered at the end of the hour from memory.

You leave with one body of material structured and the method that admits the next one.

Everyone who works with material for long enough arrives at the same problem: there is more of it than can be held, and searching it returns files rather than answers. An index is the thing that stands between those two, and almost nobody builds one deliberately.

The first hour is about what an index is for. It is not a list of what you have; it is a map of what your material can answer. That distinction changes every decision that follows — what gets a name, what gets a relation, what is worth keeping at all. We look at why the middle of a long body is the part least well recalled, why what is given at the moment of the question outperforms everything given in advance, and why an index built around subjects fails where one built around questions holds.

The second hour indexes your own set. You bring material you already keep — notes, research, transcripts, a folder of drafts, a year of correspondence — and it is given a structure here: entries named by what they answer, relations drawn between them, the whole thing shaped so a machine can be pointed at the part that matters rather than the mass. The measure of success is stated before the work: a question you could not answer this morning, answered by the end of the hour from your own material.

You leave with one body of material indexed and the method that admits the next one. Scale stops being the problem it was, because what grows is the index rather than the search.

Every practised person carries mechanics they never name — the way they read a room, the check they always run, the image they reach for. Those mechanics decide the quality of their thinking, and they were built mostly by chance.

The first hour is about building one deliberately. A mechanic is a relation named, practised across varied material until it operates without effort, and then composed with others. We look at how a skill passes from attended to free, why feedback that helps in the moment can prevent learning, and why varied practice feels harder and lasts longer.

The second hour builds yours. One relation you want to hold — a way of describing, reasoning or noticing — is named in the room, practised on material that changes each round, and then tested by withdrawing support to see what the mind does on its own. Then the harder question of composition: how two mechanics stand beside each other and which situation calls for which.

You leave with one mechanic practised and the test that shows it holds — the first piece of an intelligence spectrum that grows every time you notice yourself relating the same thing twice.

Every practised person carries judgements they never state — the check they always run, the shape they always reach for, the thing they refuse. Those judgements are the most valuable thing they own and the least transferable, because they live in a head rather than in a file.

The first hour is about writing one down. A skill is a file: a described judgement a machine can be given, with a name, a boundary, and a description whose only job is to say when it should be chosen. The description is the mechanism, not the decoration — a perfect skill that is never selected does nothing. We look at what belongs in the file and what belongs outside it, why the boundary written into a skill matters more than its content, and how a principle becomes a mechanism rather than a note you keep failing to follow.

The second hour writes yours. One judgement you carry is taken from the room and put into a file, given its description, run against real work, and then tested by removal — a line is taken out to see what that line was doing. What survives is a working skill; what does not survive was ornament. Then the harder question of composition: how two skills stand beside each other without fighting, and which decision belongs to which of them.

You leave with one skill file that runs and the test that proves it does. It is a small object, and it is the first piece of a system that grows every time you notice yourself deciding the same thing twice.

A knowledge structure is a form your next question can be received by — and its quality shows the first time you need something you know you once understood.

The first hour is verification, because a structure built on unchecked beliefs misleads the one who holds it. Every claim is sorted into three: what is true, what is false, and what remains unknown. The third is the one people skip and the one that does the work, since an honest unknown can be pursued. We look at how a belief survives the evidence against it, and what a claim needs before it earns a place.

The second hour composes. One subject of your own is taken from a heap to a structure: what you already hold is gathered and its gaps named, the relations between the pieces are drawn until the shape of the subject appears, and the whole is composed into one map with its sources kept beside it. The last part of the hour is how the next subject enters without rebuilding anything.

You leave with one subject composed whole, the map that holds it, and a method that admits the next.

A library is not a store of what you have read. It is a structure your next question can be received by — and the difference shows the first time you need something you know you once knew.

The first hour is verification, because a library built on unchecked material is worse than none. Every claim is sorted into three: what is true, what is false, and what remains unknown. The third is the one people skip and the one that does the work, since an honest unknown can be pursued while a false certainty cannot. We look at how a source is read laterally rather than deeply, why a citation should be assumed fabricated until it is opened, and what a claim needs before it earns a place.

The second hour composes. One subject of your own is taken from a heap to a structure: what you already hold is gathered and its gaps named, the relations between the pieces are drawn until the shape of the subject appears, and the whole is composed into one file with its sources kept beside it. The last part of the hour is the part that matters most — how the next subject enters without rebuilding anything, which is what makes a library rather than a document.

You leave with one subject composed whole, the map that holds it, and a method that admits the next. The value compounds: the second subject is easier than the first, and the tenth is easier than the second.