Education · the record

What is proved, and what is not

Every figure about the Learning System on this site comes from a guard of the library, named beside it, and every claim has its limit beside it. The translations are drafts, the cells run on the desktop, and no institution has adopted it. This page says each of those first, then the figures, then two tools run for you.

Three limits, said first

The translations are drafts

No native speaker has reviewed any of them yet: 0 of 35 units in French, in Arabic and in Hausa (the run below prints it). English is the original. Each translated chapter and world page opens with a note in its own language (in French: Traduction provisoire : pas encore relue par un locuteur natif) until a sign-off is recorded for it, and then names who signed. So "four languages", here and on the other pages, means four editions, three of them not yet read by a native speaker.

The cells run on the desktop

The reader is a page that opens in any browser, but its cells run on the desktop, and the page says so on every cell. Running them in the browser waits on the Softanza browser engine and is not built. Every figure below was measured on the desktop.

No institution has adopted it

A bank and a university exist in the library as two reference overlays, written to show what an overlay is. They are not customers, and this site names no institution as one. The author has not chosen a first institution.

Two more things, in plain words. The tutor is rule-based: it uses no AI model, and this site does not call it AI. And the licence is the repository's own, MIT, decided on 2026-10-05 on the author's delegation (section 8 of the charter): the author may replace it before a first institution ships.

The figures, each beside its guard and its limit

Measured by the module's authors on 2026-10-05, in a fresh checkout of the library at commit 4a184e6f0. A guard is a file of the library that runs and prints how many of its assertions held. This site did not rerun the fourteen guards, which take about twenty minutes together; it reran two tools, the learner's desk and the reviewer's sheet, shown further down.

The course

The Elementary Introduction: 15 chapters in English, French, Arabic and Hausa, which is 60 editions. Every cell ran, every promise was kept, no output is stored. Guard course_narrated(opens in a new tab), 249 assertions. Beside it: French, Arabic and Hausa are drafts, 0 of 35 units reviewed; the cells run on the desktop.

The exercises

23 exercises that prove themselves: each refuses its wrong answers and accepts its right ones, by running them. The same guard. Beside it: they are checked on the desktop; the tasks in French, Arabic and Hausa are drafts.

The worlds

3 teaching worlds (a restaurant, a cooperative, a school), a page each in 4 languages: 12 pages. Guards worlds_narrated(opens in a new tab), 26 assertions, and world_pages_narrated(opens in a new tab), 45. Beside it: the nine pages in French, Arabic and Hausa are drafts.

Skills and levels

25 skills in 7 families, and 5 levels, each earned by a project folder that passes its guard. Guards spine_narrated(opens in a new tab), 33, and levels_narrated(opens in a new tab), 35. Beside it: the skills' wording in French, Arabic and Hausa is a draft; no institution has adopted the system.

The tutor

It asks the open question, refuses to write the answer, and never explains a chapter ahead of the learner, in 4 languages; no AI model. Guards tutor_narrated(opens in a new tab), 45, tutor_gaps_narrated(opens in a new tab), 31, and slice_narrated(opens in a new tab), 68. Beside it: it is rule-based, not AI; its texts in French, Arabic and Hausa are drafts.

Institutions

An overlay (its own world, exercises, rules and name) passes a court; there are two reference overlays, a bank's and a university's; the steps of the guide are performed by a guard. Guard institution_narrated(opens in a new tab), 39. Beside it: no institution has adopted it; the two overlays are references, not adopters.

Ownership

Progress is a text file in the learner's folder, and a pass carries the fingerprint of what was run. Guards slice_narrated, above, and desk_narrated(opens in a new tab), 24. Beside it: the checker runs on the desktop.

The demo

Fifteen minutes, eight scenes, 20 claims proved. Guard demo_narrated(opens in a new tab), 17 assertions; the presenter's guide is DEMO.md(opens in a new tab). Beside it: it runs on one laptop, offline, on the desktop; the French, Arabic and Hausa it shows are drafts.

The whole

14 scenario guards, 716 assertions, run in a fresh checkout in about 20 minutes. The files are test/education(opens in a new tab). Beside it: they were run by the module's authors; this site did not rerun them.

The reader page

All of it, 15 chapters and 3 worlds (72 editions), built into one HTML page in about 100 seconds. Tool build_reader.ring, guard desk_narrated. Beside it: the page opens anywhere, its cells run on the desktop, and every draft says so.

The learner's desk, run for real

The learner's desk is a command of the library, run from its own folder with a learner's folder and a verb. Everything below was run by this site, on the library at commit 4a184e6f0, in a temporary learner's folder that is not the library's.

A learner who has just begun is on chapter 1 and has passed nothing. The desk says where they are, and what the first level still needs: four exercises and a project, and the later levels after them.

At the command line

learn.ring ada status

Output

LEARNER ada  (en, world: workplace)
ON chapter 1: find-then-apply -- Find, then apply
PASSED 0 of 23 exercises
LEVEL s0: missing 5 -- ex-01-01, ex-02-01, ex-03-01, ex-04-01, project-s0
LEVEL s1: missing 9
LEVEL s2: missing 14
LEVEL s3: missing 21
LEVEL s4: missing 24