The DREAMAX Method: Spaced Repetition + AI Conversation for Faster Japanese

One memory model for every skill (not just flashcards)

The most robust finding in learning research is simple: memory lasts in proportion to spaced, effortful recall — the spacing effect, the testing effect, and "desirable difficulty." Spaced-repetition software (SRS) turns this into a schedule, reviewing each item right before you would forget it. The modern engine for this is FSRS, which models every memory by its Difficulty, Stability, and Retrievability.

Ordinary SRS apps apply this to vocabulary flashcards only — grammar, reading, and listening are left to linear lessons with no long-term scheduling, so a grammar point you met in week two is never deliberately brought back in week six. DREAMAX lifts the unit of scheduling from "a flashcard" to "a knowledge component": one FSRS scheduler runs over a graph of hundreds of components spanning every skill — a particle rule, a reading sub-skill, a listening distinction, a grammar transformation, and each word.

Every time you finish a lesson step, answer a reading question, score a read-aloud, or use a word correctly in conversation, that becomes a review of the underlying component. The result: one honest, always-current picture of what you know and when each piece will fade — and reviews that appear exactly when you are about to forget, so you spend the fewest minutes for the retention you want.

AI conversation that actually makes you speak

Lessons build recognition; mock exams build test performance; conversation is the only place you produce free, real-time Japanese under real communicative pressure. Second-language research has long held that this — comprehensible input you negotiate in a real exchange, plus output you are pushed to produce — is what turns knowledge you "kind of know" into fluency you can use.

DREAMAX conversation is never an open-ended chatbot. Each session is a scenario with a goal (ordering at a café, a job interview, making a complaint), pinned to a JLPT level and to specific skills. The AI partner is capped so it never replies above your level, and sessions are seeded from your weakest components — so you practice exactly what you are shaky on.

Three on-demand aids remove the fear without removing the effort: tap any line the teacher says to see an instant translation and a word-by-word breakdown (save any word to your review queue with one tap); get two or three suggested replies when you freeze; and receive a naturalness note after you speak. Your audio stays private, and a post-session report turns your real mistakes into your next reviews.

A placement test that respects your time

Before you study, you need an honest starting point. DREAMAX places you with an adaptive test that uses a sequential method to reach a confident decision in the fewest questions — it stops as soon as it is sure, instead of marching you through a fixed 100-item exam.

Two principles keep it honest. It never skips you past a level you have not actually cleared, and it never pretends you know foundations you do not — so you build on solid ground, not a false floor. The result is a study plan aimed at your real weak points from day one.

Real mock exams, real diagnostics, real teachers

When your test approaches, full timed JLPT mock exams — per-section countdowns, scaled scoring, a question-by-question review — show you exactly where you stand, and the diagnostics route your remaining effort to the weakest sections instead of re-studying what you already own.

And the method is not just algorithms. The curriculum is authored by educators and the placement model was reviewed independently; founder Dr. Winnie is a PhD educator who has trained over 10,000 teachers, and Mori-sensei keeps a real teaching presence throughout. DREAMAX is a real Osaka language school, online.

よくある質問