Enterprise language learning software. Language lab software for schools and language institutes: an effectively infinite, pre-prepped, auto-graded listening library of real human voices — nothing synthesized; it is vast because thousands of real speakers recorded it, not because a machine generates it. Your teachers never prep another listening exercise, and every minute of student practice is verified in a ledger. It also works for an individual who treats himself like a school — but hey.
I built enterprise software for an enterprise of one. I needed tens of thousands of graded listening repetitions to reach C1 in Romanian — 42,000 transcriptions in fourteen months, every one recorded — and no classroom tool existed that could feed that volume or count it honestly. So I wrote one. It turns out the thing one obsessive learner needed is the thing a language school runs out of every single week: prepared listening material, and proof the students actually did the work.
Listening is the spine of all language learning — and it is the skill classrooms are worst equipped to drill, because every exercise a teacher runs must first be found, cut, checked, and graded by hand. The traditional language lab solved the room and left the library empty.
A school's results live and die on out-of-class volume, and out-of-class volume is exactly what no one can verify — until it's a ledger. Assigning "thirty minutes of listening" becomes assigning a number, and reading whether it happened takes one glance. Across a cohort, that's variance reduction: the students who quietly do nothing become visible in week one instead of at the exam; the strong ones get an unbounded ceiling instead of waiting for the class.
The honest caveat, up front: this is a volume instrument, not a curriculum. It trains the ear and the ear only — your teachers keep speaking, reading, writing, and the human parts. It is the truck, not the storefront: ugly, industrial, and packed with the one cargo every language program is short of.