How learning works here

Meaningful input first. Gradually reduce the support.

1. Set the scene. Identify roles, facts and the communication goal.
2. Notice useful language. Connect familiar expressions with new ones and understand their limits.
3. Hear a model exchange. Notice what each turn responds to.
4. Change a detail. Preserve the intent while changing a document, time or action.
5. Build a response. Select from meaningful input instead of translating a prepared essay.
6. Take turns. Attempt a reply before seeing one possible model.
7. Transfer independently. Change the facts or constraints and record your own response.
8. Recall later. Attempt retrieval before revealing the original model.

What a score means

Completion records practice and self-review. Checkpoints control sequence using your response and your own judgements. Browser transcription is not pronunciation assessment. Azure scores describe a reference reading. AI text feedback does not replace a teacher or an official exam.

Support for learners

The core material assumes common vocabulary and the ability to form simple sentences; it is not a complete absolute-beginner English course. Chinese support can help with meaning. Simple, accurate language is always acceptable.

A five-minute session compresses the routine into a situation, model, response, transfer and recall. Ten minutes adds language noticing and response planning. Fifteen minutes includes all eight steps. These are suggested allocations, not promises of instant fluency.

Review that keeps the situation

Review cards retain the original task and model sentence. First respond without looking. Then reveal the model and judge whether your meaning works. A successful recall increases the interval; needing help returns it to six hours. The sequence is six hours, one, three, seven, fourteen, twenty-one and twenty-eight days. This is a product schedule, not a personalised diagnosis of memory.