Twelve scenes, 2: One body, one receipt

A courier in Yogyakarta is stuck on Newton’s third law, and the move that frees him is meant to travel to Guatemala as a technique.

Twelve scenes is a weekly series: one scene a week, from the twelve experiments we have planned for Meta DAX. Every one is planned as a study with people, and none has run with people yet; where a model has played the people, the scene says so and the numbers are labelled SIMULATED. Each is told through one person, because the scale is easier to see in one life; the people are fictional, first names only, and the places, languages and subjects are real.

Two receipts in Yogyakarta

Ayu teaches physics by day and tutors adults by night, in Indonesian with Javanese slipping in. She has forked the published Newton course and rendered its shared cores into Indonesian without touching them. Tonight’s learner is Bagus, a motorbike courier who flies kites at Parangtritis on Sundays. He is on the third law, and stuck. “If the kite pulls the line exactly as hard as the line pulls the kite, they cancel. So nothing can ever move. So the law is wrong, or the kite is.” The model’s hints name the misconception, twice, and do not land.

Ayu tries two things that fail. She swaps the kite for a boat pushing off a jetty; Bagus nods and makes the same error. She says the forces act on different objects; he hears the words and cannot use them. Then she takes two delivery receipts from his bag, writes KITE on one and AIR on the other, and says every force goes on the receipt of the thing it acts on, and a pair is never on the same receipt. Now add up the kite’s receipt only. He writes the wind’s push and the line’s pull on it, adds them, and sees a net force. The kite’s push on the air is on the other receipt and cannot cancel anything on this one. He is through.

The session records the move as hers, not the model’s. Asked whether he will share the exchange, Bagus says yes. Ayu reviews it, names it “one body, one receipt”, and it becomes a technique in the course.

Eleven weeks later, in Sumpango, Guatemala, where giant kites are flown over the cemetery on 1 November, Marisol, who speaks Kaqchikel and Spanish and shares nothing with Bagus except kites, hits the same wall in her Spanish rendering. The tutor recognises the shape of her two attempts, finds the technique in the registry, and offers it as one way among those on record, never required: a barrilete, the wind over the hill, two slips of paper. She takes it. She is through.

Where the guide and the learner sit

Ayu is the guide and Bagus the learner, but whoever is a step ahead, right now, on this, is the guide. When the hints fail, Ayu is ahead of the model: it named the misconception and could not move him, and she did. The sequencing, the quiz and the record come off her; she reports spending the evening watching Bagus, not the page. In Sumpango she is a guide at a distance, as the educator in Manila was in episode 1, through a technique, not a rubric. The scene leaves Marisol’s seat open: she meets the wall as a learner does, yet reports, as a guide would, where her attention went. Nothing to translate, so she spent it on the kite.

What the experiment would test

This is X01, “One thread, two worlds”. The repository describes it twice. The card is the first version, simulated; the scene in the next-experiments doc is the fuller one, with people. With people it is planned, not run.

The card asks whether a guide’s good move can be captured as a technique record with no subject nouns, and reused in a different course and language. It measures retrieval when the trigger recurs, learner-advance against a control guide without it, and transfer across languages. The kill line: if it does not beat the control, or works only in its original language, technique records do not compound, and we drop them.

Three model-played passes have run, on 9, 18 and 19 October, all SIMULATED. In the first, three learners an arm, the technique cut attempts to pass from 5.5 to 2.33 and lifted the pass rate from 67 to 100 percent; its write-up says the gap was built into how the model-played learners were told to behave. The second carried captured moves between English and Indonesian, both ways, into new subjects; all six reuses landed, but took 1.83 attempts against 1.50 for a guide inventing the move fresh. The third gave a guide no technique: 1.67 attempts against 1.83, a gap of 0.17 inside a run-to-run spread of 0.67, and those guides found the same move unprompted. Both later runs are inconclusive on the kill line: a move can be carried across a language, but nothing yet shows that carrying it is faster.

The fuller version asks whether a guide’s move, made once for one learner, can be offered in another language and culture without leading either learner. It measures reuse by content hash and later learners’ attempts to pass with the technique offered against the model’s hints alone, and records human delta value on both halves: alone, the receipt move, which no hint produced; together, the technique as a shared asset, which neither could have made alone. Its kill line: if learners pass no faster, the record is noise; if they flag the offer as steering, it is a defect.

What would make it fail

The card still lists the technique-record schema and the trigger-condition matcher as missing; a proposed schema validates in simulation but is not merged. The stuck signature is an exact-match key, and whether Marisol’s confusion is the same as Bagus’s is a human judgement a key cannot make; across languages it can fail silently. Only a person can say whether the move travels without its receipts. And the record may be redundant for anyone who can find the move alone.

Two designs are already ruled out: a record that smuggles in subject nouns, which is a content node and will not transfer, and a move retrieved for every thread, which is noise. Failure is kept. The boat and the “different objects” sentence stay as marked dead ends under the same signature, and a technique that fails for several later learners is demoted to a dead end, never deleted.

Why it matters at scale

Every learner who reaches that misconception after Marisol, in any language, is offered a move made by a person who was in the room, as one way among others. Ayu’s failed moves compound too: the next tutor is told what did not land. How knowledge compounds under a constraint follows this scene through the record. The card calls it the smallest unit of compounding intelligence: not facts, but moves. It feeds X02 and X12.

Try to break it

This is written precisely enough to be argued with. If you can think of a move that would not travel, a technique offered when it should not be, or a learner it would lead, send the case to Jason on LinkedIn or open it as an issue on GitHub. All twelve experiments are here: https://github.com/daxfoundation/metadax/blob/main/docs/NEXT-EXPERIMENTS.md