If you are the tutor in the group

In a group of families there is usually one person the others turn to when the maths gets hard. If that is you, you are a guide: whoever is a step ahead, right now, on this, holds the seat, the same one a parent holds at the kitchen table. Build a kit once, for the topic you dread teaching. Then remake it as an Edition for each learner you have, in the world they love, with the maths left exactly where it was. The parent who was not in the room gets a report they can read in a minute.

One kit, then an Edition for each learner

Open Words into Algebra. It began as an interview: a maths tutor who mostly taught fifteen- and sixteen-year-olds in England was asked what learners find hardest, what he found hardest to teach, and what he wished he had. The hardest topic was turning a sentence into an equation; the hardest problem was the clock. The kit has a page for the tutor and a page for the learner, five sentences (“Double last season’s goals, then add 5 cup goals, and you get 23” becomes 2x + 5 = 23), five scenes you can press, and three routes. Before every example the learner says the sentence back in their own words, and you wait. That pause was the tutor’s own idea, and the page says to cut an example before you cut the pause.

Football is only the setting. In Editions meet Nova’s Edition: Nova is fifteen and into space. The five equations, the six named slips, the three routes and the pause are the same. The balance sentence, with scarves, a badge and a programme in the football kit, now has mission patches, a star map and a telescope filter, and still comes out as 2x + 4 = x + 11. The tutor’s page is rewritten around who she is; one line reads “Don’t explain space to her: let her correct you.” For another learner you change one line, which the page gives you: “[age], loves [their interest], and [one thing about how they learn].”

The part parents see is the report. The Tutor’s Desk follows an invented tutor, Nadia, and an invented learner, Theo, who is twelve, through one session: her rough notes, the report she writes for his parent, and a free coaching prompt about the problem she keeps hitting. The report has four short parts (what happened, what worked, the misconception she saw, one clear next step) and starts from what Theo can do. His fraction sums are secure; what stops him is turning the words of a problem into the sum to do, the step Words into Algebra is built around. Obsidian Delta’s sample report shows the same four-part shape beside the rough notes it came from.

What the evidence supports, and what it does not

Our own evidence is a design, not a result. Thirty to One counts thirty seats behind one learner: the mistake-mapper, the storyteller, the note-taker for next time. A kit fills many in advance; the post is plain that some are only on paper so far. The seat that is not a task is yours: noticing the learner has gone quiet, deciding whether to push today or leave it until tomorrow. The Words into Algebra page gives it a card, What only you can do: the kit holds the order, the examples and the routes, and cannot tell the silence of thinking from the silence of being stuck. That is human delta value; the building the kit takes off you is reclaimed cognition, to spend there. Nothing in Meta DAX has been measured with real learners yet.

Outside evidence points at the same seat. In Tutor CoPilot, researchers at Stanford University randomly gave some tutors in a U.S. school district’s virtual maths tutoring an AI tool that suggested responses as they tutored; the tutors chose what to use. Their learners, in grades 3 to 6, were 4 percentage points more likely to pass the exit ticket that lets them move on to the next lesson (62% to 66%), and 9 points more likely with lower-rated tutors (56% to 65%). The limits are in the same paper: the measure was the exit ticket, the authors found no statistically significant improvement in end-of-year maths test scores, and the tool may help experienced educators less. It is a live chat helper, not a prepared kit, so we cite it for where it points the AI, at the person in the guide’s seat, not because Meta DAX has shown such gains.

What to do this week, and where to go next

Name your hardest topic, the one you dread teaching, and answer the interview’s three questions for it. Open the kit and read the tutor’s page first; it is yours to check, because you are the one who can. Then pick one learner and decide what you would change for them. The tutor door and the any-AI guide show how to run it on the AI you already have. Meta DAX is free. It is built on paid compute.

This closes the series. It began with the argument that where your child is stuck is the whole brief, and a kit starts from the same place: the topic your learners find hardest. Session two, the part before this, is about the note that carries one session into the next, and the algebra kit ends with one for you to paste in next time. If tutoring is your work, Obsidian Delta’s post on a tutoring practice with a kit for every hard topic treats it as a practice, and the company sells the same building as done-for-you work.

Which topic do you dread teaching, and what does the learner in front of you love? Answer in a comment wherever you found this, or in an issue on the metadax repo.

Sources

• Wang, R. E., Ribeiro, A. T., Robinson, C. D., Loeb, S., & Demszky, D. Tutor CoPilot: A Human-AI Approach for Scaling Real-Time Expertise. arXiv:2410.03017, v3, 10 October 2026 (first posted 3 October 2024). Stanford University. https://arxiv.org/abs/2410.03017