Mastery, fluency, and a thermometer that respects autonomy
Rodybee is built on a few non-trivial pedagogical decisions. Here is what every one of them is, and why it's there.
1. Skill trees, not flat lists
Every program (arithmetic, reading, geometry, fractions) is a tree of small, single-target skills with explicit prerequisites. A kid can't 'guess their way up' β the next skill literally cannot be opened until the prior ones hold.
2. Dual metric: accuracy AND fluency
We don't accept 'correct but slow' as mastered. A skill is mastered only when accuracy passes a threshold (typically β₯95%) AND average time per exercise is at or below target. This kills the 'I knew it but I couldn't keep it in my head' gap.
3. Spiral review built in
Every practice session is 80% from the current hex and 20% from previously mastered skills. Without this, mastery quietly evaporates over weeks. With it, what was learned stays learned.
4. The readiness thermometer
Most platforms either force a test on a schedule (anxiety) or never test at all (no closure). Rodybee uses a continuous multi-signal check β every-skill mastered + last 10 strong + at least 24h since the last test + at least 3 distinct sessions in the block β to decide when to SUGGEST the block test. The kid can say 'maybe later'. If readiness keeps holding for many sessions, the test becomes mandatory in a 3-day window the kid chooses.
5. Age-appropriate framing
For kids 8 and under, we never use the word 'test' β it's a 'challenge' or a 'reto'. From 9 up, 'test' is fine. Tiny copy change, big difference in how the kid relates to checkpoints.
6. The parent has real control
From the dashboard, a parent picks how insistent the app should be (permissive / balanced / strict), or can force a checkpoint themselves when they sense the kid is ready but stalling. The kid keeps autonomy; the adult keeps oversight.
7. The block test is a moment, not a verdict
Pass (β₯85%) and the next hex unlocks. Don't pass and the system marks the weak skills, sends the kid back to practice them, and tries again later. There is no 'failed' label, no public score, no penalty.