Begin with a real context
The learner names a topic or situation. The activity must remain recognizably about that request rather than inserting it into a generic template.
Learning methodology
How Cashi turns context into focused English practice
The method combines a learner’s real goal with CEFR-shaped difficulty, active production, evidence-based feedback, and review. This page explains the system without treating AI output as infallible.
Shape the task to CEFR
Difficulty changes through vocabulary, sentence structure, length, support, and the kind of thinking required—not only playback speed or word count.
Require active production
Speaking and writing ask the learner to produce language. Reading, listening, grammar, and vocabulary require retrieval instead of passive exposure alone.
Ground feedback in evidence
Feedback should refer to the learner’s answer, transcript, selected option, or measured speech signals. Unsupported praise and generic advice are rejected.
Return the right material
Saved vocabulary and skill history can return for review. Difficulty can become supportive, balanced, or challenging as new evidence appears.
Keep the learner in control
Learners can choose a level, change topic, request another activity, review results, or continue without treating one score as a permanent judgment.
Limits and quality checks
AI-generated material can be wrong, generic, culturally awkward, or misaligned with the requested topic. Generation routes therefore use structured outputs and validation contracts, and important flows have regression tests for language, level, topic relevance, answer support, and readable feedback.
The adaptive assessment provides a working level for practice. It is not a substitute for an accredited CEFR examination, a teacher’s full evaluation, or professional advice.