# teach-concept Examples

## Example 1 - Beginner Concept Intro

### User Prompt

"Can you explain gradient descent? I keep hearing it in ML tutorials."

### Expected Behavior Pattern

- simple definition and intuition,
- one numeric or visualizable example,
- one understanding check,
- one tiny implementation exercise.

### Anti-Pattern to Avoid

- immediately deriving full optimization equations without calibration.

## Example 2 - Intermediate Depth Expansion

### User Prompt

"I know basic gradient descent. Explain why Adam can converge faster in practice."

### Expected Behavior Pattern

- acknowledge prior knowledge and skip basics,
- compare update behavior and practical tradeoffs,
- include failure case discussion,
- ask transfer question involving optimizer selection.

### Anti-Pattern to Avoid

- repeating beginner explanation verbatim.

## Example 3 - Failure-Recovery

### User Prompt

"I still don't get what overfitting means."

### Expected Behavior Pattern

- diagnose misunderstanding,
- switch to simpler example and contrast train vs validation behavior,
- ask learner to predict what happens when model complexity increases,
- confirm corrected understanding.

### Anti-Pattern to Avoid

- restating the same definition with different wording only.