Scratch AI Lab camp plan
Use this page when running Code Creators: Scratch & AI Lab as a 5-day camp.
At a glance
| Day | Blocks | Goal |
|---|---|---|
| 1 | Setup, Stage 1, Stage 2 | Students train a first image model and understand examples. |
| 2 | Stage 3, Stage 4 | Students test the model and connect it to RAISE Playground. |
| 3 | Stage 5, Stage 6, Stage 7 | The AI controls a sprite and the game gets rules. |
| 4 | Stage 8, Stage 9, catch-up | Students customize, rehearse, and prepare demos. |
| 5 | Stage 10, polish, parent demo | Students explain both the game and how the AI works. |
Minimum viable finish
A student should leave with a trained model, a Scratch/RAISE project that responds to predictions, and a short explanation of what examples the AI saw.
Coach triage
- If training is slow, shrink the dataset before adding features.
- If predictions are unreliable, improve lighting and add clearer examples before debugging code.
- If a student is behind, skip extra customization and focus on one working control path.
- If a student is ahead, send them to Stage 10's explanation and demo polish work.
Common stuck points
- Students mix up training examples and testing examples.
- Browser permissions can block camera access.
- Learners may think the AI is "thinking"; keep redirecting to examples, patterns, and confidence.