You don't memorize rules. You grow a language network.
Each of the five stages maps to one mechanism from second-language acquisition research. Take it in (XRI), put it out (US), and every cycle settles one more form-meaning construction into place.
You meet comprehensible input in a low-pressure setting. The affective filter drops, so understood input can feed acquisition.
Filling the blank to pull the sentence out is retrieval — shown to fix memory more firmly than rereading.
Typing the expression yourself is your first act of production. As Swain's Output Hypothesis holds, making it yourself surfaces the gaps and turns understanding into usable language.
Shadowing the line the moment you hear it encodes sound and motion together, building fluency that rolls off the tongue.
In a conversation where only the situation is given, you produce learned expressions on your own — pushing toward automaticity.
We won't promise "fluent in a few weeks." What XRIUS shows is measured improvement, and only that. The idea that AI learns patterns from input to build a language network is an analogy, and only an analogy — people learn from meaning and context (comprehensible input is the decider); AI handles it statistically, without "understanding." Saying they're the same would be false, so we don't.
One expression is one line of your language brain
Language isn’t a pile of rules you memorize — it’s a network that emerges from real use. Every expression you take in and put back out becomes one line connecting two points in your language brain.
Lines pile up, points start linking to points, and the web gets denser. That growing density is what we can actually measure — not a vague feeling of “fluency.”
An analogy is just an analogy. The “language network” is a way to picture it, not a brain scan — what we claim, we measure.