Most educational apps don't work. Not "could be better." Not "work...

If a task can be completed WITHOUT using the skill you're teaching, some learners will find that other path.
Example: A reading app shows "elephant" next to a picture of an elephant. Child taps correctly. But did they decode or just match the image? The app can't tell. The dashboard shows progress. For all intents and purposes, the learning never happened.
Prerequisites aren't optional. Learners who advance with gaps will struggle with everything built on those gaps. This looks like "struggling with maths." It's actually "moved forward before ready."
But session time ≠ learning time. An hour of clicking, watching, and recognising can involve almost zero actual cognitive application.
Because compliance conflicts with the metrics apps optimise for:
- Session time (effort = fatigue = shorter sessions)
- Completion rates (mastery gates = people get stuck)
- User satisfaction (real feedback feels harsh)
- Can tasks be done without the target skill?
- Multiple new things at once?
- Can learners just skip ahead?
- Watching or actually producing?
- Real feedback or just "try again"?
- Does practice match the real performance?
Most apps fail most of these checks imo. That's why most apps don't produce reliable learning.
- Engagement
- Personalisation
- Smooth user experience
- Motivation
- Time-on-task
These conditions are instructional invariants: the non-negotiable constraints under which learning reliably occurs. You can innovate endlessly within them. You cannot design around them. 🧵