Built for the whole curriculum.
Every subject and every grade, structured against a country's own national standards — not against ours.
Atlas is built on the national curriculum of the country a child lives in: each competency is broken down into the smallest units a child can get right or wrong, those units are ordered by what each one depends on, and the mastery estimate is kept on them.
Why not just translate an American curriculum?
Because it produces a product that teaches the wrong things in the wrong order and calls it localisation. Most learning software is built to US standards and then translated. That is backwards.
A Peruvian child is assessed on Peru's national curriculum; a Mexican child on Mexico's. They are not the same document, they do not put the same ideas in the same years, and they do not call things by the same names. Atlas is built the other way round: the structure comes from the country's own standards document, and the content is authored to it.
What counts as a "skill" inside Atlas?
The smallest unit a child can actually get right or wrong — and that unit is what the mastery estimate is kept on.
A national standard says something like "solves problems involving quantity." That is a competency: it describes where a child has to end up, not something a tutor can teach on Tuesday. Atlas decomposes it until what is left can be observed one piece at a time, then arranges those pieces by dependency — which is what lets it start a child at their actual gap rather than at their birthday.
| What it is | What Atlas does with it | |
|---|---|---|
| The competency | What the national document says: "solves problems involving quantity." | The starting point, never the unit of teaching. |
| The skill | The smallest piece a child can get right or wrong. | This is what the mastery estimate lives on. |
| The order | What each skill needs underneath it before it can hold. | Lets Atlas start at the real gap, not at the expected grade. |
A machine may propose a mapping. Only a person may approve one.
This is a rule we would rather state than be caught not having.
Generating a mapping between a skill and a national standard is easy and fast. Being right about it is neither. So Atlas's rule is that a machine may propose a mapping and a qualified specialist in that country's curriculum must approve it — and until they have, we do not say "aligned", we do not say "covered", and we do not publish a coverage figure.
Plenty of products claim alignment to a national curriculum on the strength of nobody having checked. We would rather have the slower claim and have it be true.
Where it starts.
Mathematics, from the earliest years. It is the subject where the prerequisite structure is clearest, where a child can be assessed without reading a word, and where a wrong answer tells you the most about what they were thinking.
Language, science and the rest are built on the same engine and the same lesson formats. The engine does not care which subject it is sequencing — which is precisely why "everything, in one" is a build problem rather than a research problem.
Does Atlas listen to my child read?
Not yet, and we will not fake it. Assessing whether a child can read means listening to a child read aloud, and doing that well for a five-year-old in their own accent is genuinely hard.
Atlas is architected for it and does not pretend to do it yet. Where reading appears today, the child answers by touching rather than speaking — and we say which is which instead of leaving it vague.
Every subject, every grade, in one thing.
One email when Atlas opens. Nothing else.