Curriculum as Data, Rendered Anywhere
A curriculum that no single app owns
Everything a child at Fido reads and practices is written down as data: every topic, every explanation, every exercise. This article explains what that means and why we do it. It covers how topics are linked, what is attached to each one, how one written piece appears in several places, how the curriculum lines up with the Bavarian LehrplanPLUS, and who checks it. It is written for parents, educators, funders, and anyone who might write for it.
It is the second article on Fido's software. The first, The Shape of Fido's Software, names the parts; this one takes the curriculum further. The basic model of a topic is introduced in Seven Kinds of Knowledge and the Curriculum Map, and this article does not repeat it.
As before, we say what runs today, what is planned, and what is still undecided. The numbers are counts from the curriculum in November 2026, rounded.
Why Data, Not Pages
There are two familiar ways to hold a curriculum. One is pages: a textbook, a worksheet, a website. People can read them, but software cannot tell which page a child needs next. The other is content built into an app, where each lesson is programmed as a screen. Our first interactive articles were made that way. Only that one app could show them, and every new article needed a programmer.
We now write the curriculum as data. Each topic is a short text file in
a fixed form that says what the topic is and what it builds on. Its
explanation is a piece of a web page, and its exercises are a list.
Nothing in these files belongs to one app. Together the topics form the
Writing it this way buys four things. The same piece appears on the tablet and on the public explorer without being written twice. A machine can check every piece before a child sees it. A reviewer can read each change line by line, as with program code. And the curriculum can be published under an open license for other schools. That is the plan; none of it is published yet.
The price is a stricter way of writing. An author works in files of a fixed form, not in a word processor, and a broken link stops the build. We accept this because thousands of topics stay correct only if a machine checks them on every change.
Topics and the Links Between Them
A topic is the smallest unit of the curriculum: adding within 20, the
accusative case, turning left on a bike. The
Builds on. A prerequisite says that one topic must be secure before another begins. It is the only link the mastery engine is meant to use to decide what a child may practice next; for now the engine follows the order of a course instead, as The Mastery Engine explains. About 2,800 topics name at least one prerequisite, with about 4,600 such links in all.
Part of. A topic can belong to a larger one, as a skill belongs to a group of skills or a section to an article. This link draws the map that children and guides see. It never holds a child back.
See also. A topic can point to a related topic, often in another subject. It is a pointer for the reader, nothing more.
The three are kept apart on purpose. The order in which material is read is not, by itself, a prerequisite. One topic is made to wait for another only where it cannot be learned without it. If reading order were turned into prerequisites, the mastery engine would hold a child back from things they could already learn.
Every link is checked whenever the curriculum is built. A link to a topic that does not exist is an error, not a quiet gap. Prerequisites may never run in a circle. And no topic may float on its own: each one either builds on another topic or belongs to a larger one.
What a Topic Carries
A topic on its own is a name and its links. What a child actually meets is attached to it, and five kinds of content matter most.
Exercises. Short items whose answers the software can check: choosing an answer, typing a number or a word, filling a gap, putting steps in order, recalling a flashcard. There are more than 40,000 of them, in some 2,850 sets. Most are checked on the tablet at once. A few kinds, such as a translation or a short answer in the child's own words, are judged with the help of AI.
Articles. An explanation, often with a figure the reader can move. A longer article is a case: a starting situation followed by sections, each of which is a topic in its own right. About 270 articles exist, and you are reading one of them.
Worked examples. A solution shown step by step on a figure, with spoken narration, before the child tries one alone. There are 48 so far, in nineteen subjects.
Reader texts. Texts to read inside the app. A child can tap a word to see what it means or listen to the text read aloud. There are about 260, most of them in German and the rest mainly in Ancient Greek.
Tests. A test is written as data too; we call it an instrument. Its file states what the test is for, where its items come from, and which rule turns answers into a result. Every answer is recorded, and the result is computed from those records, so it can be worked out again later. Two tests have this form so far: a reading test and a vocabulary check. The others still use older forms and are to move over.
One Piece, Shown Everywhere
The files are not yet pages. What turns them into pages is the shared engine named in The Shape of Fido's Software: one library of code that knows how to draw an article, an exercise, a worked example, or a reader text, and how to check an answer.
The same library runs in two places today. In the app, it draws every article, exercise, and reader text a child sees on the tablet. On the public explorer, it draws the same pieces for anyone with a browser. A correction made once therefore appears in both. The app itself holds no lesson of its own. It decides which piece to open next and keeps the record of what the child did.
Interactive figures are reusable building blocks: a graph whose point a reader drags, or a clock that converts between 12-hour and 24-hour time. The library holds about ninety building blocks, most of them figures like these. An author places one in an explanation with a single line and sets it up with a few settings. So a new article needs writing, not programming, unless it calls for a building block that does not exist yet.
Some work is done in advance, when the curriculum is built. Formulas are typeset then, so they appear at once. Articles for the tablet are prepared then too and travel with the app, so they open without a connection. A new article needs no new app code, but it reaches the tablet with the next update of the app.
The build also refuses to finish when a page would come out wrong: when an explanation file is missing, or when a highlighted term has no color assigned.
The Curriculum Tree and LehrplanPLUS
The
Every curriculum has the same five-level tree.
- Subject, such as German or mathematics.
- Stage, a binding package of skills, called a Stufe in the Fido model.
- Area, a few courses that belong together, drawn as one screen of the course map.
- Course, an ordered list of topics, articles, or exercises that a child works through.
- Topic, the unit from the graph.
LehrplanPLUS is written in the same form, one tree for each of the state's subjects. There the first level is the type of school, the stage is the school year, and the area is the state's own learning area. Each entry names one of our topics, the passage of the state curriculum it covers, and the state's wording of the skill. Twenty-two state subjects are mapped so far, referring to about 1,460 topics. The figure follows one topic into two of these curricula.
The accusative case shows how this works. It is written once, with one explanation and 25 exercises. The course in German as a foreign language places it in its first grammar area. The Bavarian curriculum places it in year 3, where children learn to recognize the four cases. A child who masters it in one course has covered it for the other, and the data can show which part of the state curriculum that is. A report for the school authority built on it is planned, not built.
Who Writes It, How It Is Checked, What Is Missing
Fido's team writes the curriculum, with AI drafting much of it and people checking it. How we use AI is told in How We Build the School and in How a Small Team Builds Fido's Software with AI. Teachers and subject specialists can contribute too. In the first courses set up for it, a contributor asks for a change in plain words, sees the new version beside the old one, can try it with one group of children, and then hands it in. A person approves every change before it joins the curriculum.
Checking happens in four layers. Machine checks run on every change, on links and files alike. Review marks say how far an article has come: in progress, ready once it passes our editorial checklist, or verified once a human reviewer's check is recorded. Release stages say who may see a course: draft, testers, school, or public. A group of children sees a course only once its stage includes that group. Reports from use come from testers, who flag problems from inside the app and the explorer.
What exists today: about 4,400 topics, some 2,850 of them with exercises and about 1,350 with an explanation of their own, most in both German and English. Of the articles, 224 are marked ready and three are verified. Almost every course with a release stage is open to testers only, and none yet to the public.
What is thin or missing:
- About 1,800 topics have exercises but no explanation yet.
- Subjects differ widely. Mathematics has about 1,250 topics and German about 770, while many subjects have only a few dozen.
- Articles that branch, and comparisons across several articles, are designed but not built.
- How much of a test's result a child should see is undecided; today it differs from test to test.
The next article in the series, The Mastery Engine, follows an answer through the engine to the next step.
Questions, objections, or keen to write for it? Write to us.