EleMate Every child is different

How it works

Five steps, and a check at every one

There is no clever trick in here. The interesting part is what the code refuses to do with what it has been given.

An illustration: a parent holding a small child's hand on one side, a teacher holding an open book on the other, and an elephant standing between them.
  1. An adult opens the record

    A child under 13 cannot consent to a service like this, so the consent is their parent's. An adult signs up, and creates the child. There is no child sign-up anywhere in the product, and no screen asks for a child's email address.

    That is structural rather than a policy: a child is not a row in the accounts table, so there is no address column to fill in, nothing to email them, and no password reset that could ever reach a child directly. A first name and a date of birth is enough to open a record. Everything else gets asked at the moment it would help — never as a forty-question wall.

  2. You, their teacher and they themselves fill it in

    You add results, homework and what you noticed. Their teacher adds assessments, observations and feedback. They say how a task went, mark their own work and set goals. Nobody has to know everything, because between you, you already do.

    Marks are stored honestly: a score and what it was out of, both optional, neither implied by the other. Seven means nothing until you know whether the paper was out of ten or out of forty. Age comes from a date of birth, so the year group moves up on the first of September by itself rather than quietly being a year out.

  3. A review reads that child's records, and only that child's

    Every record carries a short reference — A12 for an assessment, H5 for a piece of homework, N3 for a note, T7 for something noticed about how they work. The review is built from one child's references and nothing else.

    That is also the isolation boundary. The set of references handed to the model comes from one child's rows, so no other child's data is reachable from inside a review — not by accident, and not by a clever prompt.

  4. Every finding is re-checked before anybody reads it

    The model is asked; the code decides. Each finding has to cite its references, and each reference has to resolve to a record belonging to this child. Anything that names a condition is refused. Anything that puts a limit on the child is refused. A development area has to rest on an assessment or a piece of homework — a note about how somebody works can shape a strategy, but it can never on its own become a claim about what they cannot do.

    A review is never silently empty, either. If nothing has been recorded, or every finding was refused, or the model could not be reached, that is stored in words with the reason — because four refusals out of six is not a clean review, and hiding it would make a thin record look like a confident one.

  5. Your child opens a different page from you

    Same record, two lenses. You and their teacher see development areas and the evidence under each. They see what to do next and what has helped. A development area is not sent to a child's app at all, so it cannot be revealed by a bug in a screen.

    Holiday work follows the same instinct. No more than two activities in five may aim at a development area, and when a pack goes over that line the extra remedial activities are dropped rather than padded out with filler. Six weeks of drilling a weakness is the pack a child quietly stops opening.

What you actually see

A finding, with the records it came from

A claim, its evidence, and something to try. Each reference opens the record it names, so you can read the original and disagree with us.

Invented, to show the shape

“Fractions have gone wrong three times since February, and each time the questions were written as word problems. The same operations set out in figures were answered correctly on the same days.”

Drawn from

A12 A19 H5 N3

Something to try. Read the next word problem aloud together and write the sum out before anybody attempts it. If that is the difference, it will show up in the next two pieces of work — and if it is not, the record will say so.

Notice what is not there: no condition, no “ability”, no comparison with the class, and nothing about a type of learner. There is a pattern, the records behind it, and a suggestion for the adult who knows the child.

And then they move

What happens when they change school

New teacher in September, new school at eleven, a college after that. Each time, the adults in front of your child begin again from a name, a set of grades and whatever the last lot remembered to pass on.

Because the record belongs to the family, it is already there: three years of evidence, and what has already been tried, attached to the claims it supports. A school gets access while it teaches the child, and that access ends when the relationship does.

An illustration: an elephant handing a small card to two adults standing together.