NoteThe operating system of an honest school — from the silent baseline on day one to the closed-loop intervention by month three.
We do not begin with a stressful diagnostic. The first three days are stealth — five short, game-shaped exercises woven into existing classwork. By Thursday we have a high-fidelity map of every child's prerequisite knowledge, without a single child having sat what feels like an exam.
Bayesian Knowledge Tracing · IRT-weighted diagnosticsEvery interaction the child has with the platform — every quiz attempt, hesitation, abandoned question — is captured per concept. We do not grade the answer; we grade the thinking process. The platform integrates with the school's existing LMS or runs on a lightweight tablet that the school keeps in the classroom.
Event telemetry · 200ms event resolution · concept-level rollupsEach Sunday night, the engine runs. It distinguishes a careless slip from a conceptual gap. It predicts each child's likely forgetting events for the next two weeks. It identifies which clusters of children share a misconception that the teacher could address in one well-placed lesson rather than five small ones.
Forgetting-curve regression · concept-graph clustering · cohort comparisonThe teacher receives a Monday Brief — three children to watch this week, two concepts to reinforce in class, one suggested intervention. The parent receives a weekly note in plain language. The principal receives a monthly health view. None of them is more than a page; all of them are specific.
Natural-language generation · role-tuned templates · plain-language proof readEvery intervention is tracked. If a suggested practice set worked, the system promotes it for similar students. If it didn't, the system tries a different shape. Over a term, the engine learns what works for this specific school's students — not in the abstract, not for averaged children, but for the actual ones in the actual rooms.
Reinforcement learning · per-school priors · monthly retrospectivesOrderThe order matters: schools commission, parents notice, teachers run, students learn. We organise the system accordingly.
Operate on what students actually understand.
A principal stops running on rumour. Where retention is collapsing in a cohort, where teaching is unusually effective, where intervention is overdue — all of it becomes visible at the institutional level, monthly. Hiring conversations, board-presentation conversations, and parent conversations all gain evidence behind them.
Replace guesswork with a weekly map.
A note arrives every Friday. It says, in plain language: this is what your child mastered this week, this is what they're forgetting, this is what they're avoiding. It does not say more than it knows. It does not ask you to interpret data. It tells you what to ask your child at dinner — and lets you choose whether to.
Reclaim the hours you lost to grading.
The bookkeeping leaves your evenings. You walk into class on Monday morning already knowing who needs you and on what concept. The engine handles the pattern recognition; you handle the human being. Most teachers we have spoken to say this is the part of the job they came in for.
See your own learning, finally.
The map is theirs. The same map the teacher and the parent see — no more, no less. They watch their own retention curve. They watch the question they didn't attempt last week disappear from their map this week, because they tried it. The exam stops being a verdict on who they are.
☞ First three pilot cohorts: free. Always.
Begin a pilot ↗