Theory
The sweet spot of difficulty.
Inlet's job is to right-size the next step. Too little novelty and the hour is empty. Too much and people leave. Inlet Language is the first place we can count that.
A classroom of thirty people gets one lesson, so some of the room is bored and some of it is lost. Software does not have to pick one grade for everyone. It can pick a size for this person, today.
What Language actually measures
For each text in Inlet Language, we score how many running words are still unknown to you. The default band is readable: enough new words to grow, few enough that you can finish the page. A wall of unknown words burns the hour on decoding. A page with nothing new leaves the known set unchanged.
That count moves. Yesterday's stretch is a warmup once those words stick. A printed syllabus cannot follow you. The feed can, because it re-ranks when your known set changes.
What this is for, beyond Language
The same question will apply to anything Inlet ships: what this person can already do, what the smallest next step is, and what will slip if we wait a week. Language has a clean sensor in known versus unknown words. Other tools will need other sensors. The job stays the same: keep the hour in the band where work actually changes you.
In Inlet Language, a session looks like a story with a quiet unknown-rate chip, tap-to-define under your thumb, and aligned audio for the sentence you are on. The ranking is the part you should not have to do by hand.
What kind of work the task is asking for is a separate question. This essay is only about size. Why unknown words were a count we could ship first