A qualification says where you studied. Not what you learned.
So every time a student crosses a border, a human being reads their transcript and guesses. Learning Passport makes the guess unnecessary.
An administrator in Texas has no way of knowing what Matematika 1 in Serbia covered.
So the work is done by hand, one student at a time, and paid for twice — in fees, and again in credit that evaporates.
A course title tells an admissions officer nothing. So they estimate.
Three separate unknowns per subject — what the course actually covered, what the grade means, what it is worth — each one a judgment call by someone who does not read the language. With the curriculum already mapped, none of them is a judgment call.
РЕПУБЛИКА СРБИЈАСВЕДОЧАНСТВОгимназија · I разред
same as Algebra 1?
РЕПУБЛИКА СРБИЈАСВЕДОЧАНСТВОгимназија · I разред
Nothing about the student's paperwork changes. What changes is that the curriculum behind it has been read once — so the course opens into what it actually taught, the overlap with the receiving system is a measurement rather than an opinion, and where there is genuinely no equivalent the record says so instead of guessing.
Underneath is one claim, made once per curriculum.
That 87% is not a similarity score on two course titles. It is the outcomes of both curricula, read in their own languages and matched one by one — like this pair, which two education systems wrote a decade and a continent apart.
решити систем две линеарне једначине са две непознате
solve systems of linear equations in two variables
Two education systems that share no common language, no common notation and no common document format — and the machine can say, with the source line attached, that they are teaching the same thing. A school issues the passport. A university on the other side of the world reads it directly, with nothing to interpret and no one to pay for interpreting it.
This is not a slide deck.
The engine has been run end to end on a real national curriculum — Serbian first-year secondary mathematics, in Cyrillic, taken straight from the government gazette — and matched against the US Algebra 1 curriculum.
One subject. Two countries. What stands between that and a usable standard is no longer invention — it is scale, and scale is what costs money.
Nobody pays to cross a bridge that reaches halfway.
A passport readable against three curricula is a curiosity. Against fifty it is infrastructure. The map has to exist before it is worth anything to anyone — which is exactly the part that cannot be sold in instalments.
They pay a human per student. We pay once per curriculum.
$150–300 per student — and an evaluator for every single one.
That is the shape of the entire credential-evaluation industry. Revenue per student is decent, but so is cost per student, and growth means hiring.
Map a curriculum once. Every student who studied it is covered.
Serbian secondary mathematics is read, structured and matched a single time. After that, each student who carries it costs a lookup.
Revenue scales with students. Cost scales with curricula. There are only so many curricula in the world.
Anyone can issue a certificate. The map underneath is the asset.
Every curriculum added makes the next comparison sharper. Every school that joins makes the next school's decision easier. The market is already consolidating around ownership of academic records.
We aren't selling software into a standard. We're building the standard.
Curriculum as source code A curriculum document is parsed into its real structure, then reduced to the smallest independently teachable claims it makes — each one an action applied to an object under stated conditions.
Original language throughout Nothing is translated during extraction. Translating at that stage silently destroys the distinctions we exist to measure; cross-language meaning is carried as auditable metadata alongside the original, never in place of it.
Auditable by construction Every stage is isolated and every output stays traceable to the line it came from, so a wrong answer can be located rather than argued about — and any stage that can be deterministic code rather than a model call is code. That is what makes the output stable enough to score against a fixed answer key at all.
Matching Comparison across systems runs over those structured units in a shared semantic space rather than over course titles or free text, with genuinely ambiguous cases surfaced for expert review instead of being quietly resolved by a model.