Diagnose and choose one target
The mentor uses recent work, explanations, checks, and prior attempts to identify a narrow bottleneck. It avoids turning every session into a broad curriculum review.
See the pedagogical model behind AI-led frontend interview courses and practice: diagnose, teach, model, practice, feedback, retry, and transfer.
The learning experience follows a managed loop. It begins with what the learner can currently demonstrate, targets one useful gap, and requires the skill to survive a new context before it is treated as progress.
The mentor uses recent work, explanations, checks, and prior attempts to identify a narrow bottleneck. It avoids turning every session into a broad curriculum review.
A short explanation gives the learner a usable mental model. The mentor can demonstrate an answer or decision process, then returns responsibility to the learner through a focused drill.
Feedback points to a specific observable behavior. The learner retries the skill with less support, and completion requires a meaningful improvement rather than passive exposure.
The same skill appears later in a different prompt, project story, coding task, or mock interview. Spaced review keeps the preparation plan responsive to evidence instead of elapsed study time.