Mastery you can inspect.
Rheticus never tells you that you are 83% proficient. It shows which of five states each concept is in, and the attempts that put it there.
The five states.
One mark for each, used everywhere in the app. Shape carries the meaning, so it never depends on colour.
Not encountered
You have not worked through this concept yet.
Earned by: Nothing. Every concept starts here.
Does not count: Having heard of it. Having it on a roadmap. Having a book about it.
Understood with help
You can follow an explanation or solve a guided example.
Earned by: Completing a guided explanation and answering its check questions. Hints are allowed.
Does not count: Reading alone. Watching a video. Asking a question and getting an answer.
Practised
You have done exercises yourself, and you still need support.
Earned by: Passing two of your last three practice attempts on the concept. Hints are counted.
Does not count: One lucky pass. Exercises you only read the solutions to.
Demonstrated
You can solve an unfamiliar problem, or produce work that is assessed independently.
Earned by: Passing two of two or three tasks you have never seen, with no hints, at the rubric's independent level.
Does not count: Repeating an item you have already attempted. Any attempt that used a hint. One lucky pass.
Retained
You can retrieve the concept and apply it after a delay.
Earned by: Passing a review at least seven days after you demonstrated it. No hints.
Does not count: Still feeling confident. Recognising it when you see it.
Six rules the app enforces.
They are plain application logic. No model decides whether you know something.
Evidence only
A state changes because an attempt was recorded and judged. There is no button for marking something as known. You can always move a state down yourself.
Exposure is capped
Reading, watching, importing, highlighting and asking can take a concept to understood with help, and never further.
Hints are capped
An attempt that used a hint cannot demonstrate a concept or show it is retained.
Unseen means unseen
An item counts as new only if you have never attempted it or any of its variants.
One step at a time
A single attempt raises a state by one step at most. The challenge route is the only exception.
The challenge route
If you already know something, take two or three unseen items with no hints. Pass two and the lesson goes straight to demonstrated.
Verified, or provisional.
A model can mark work wrongly. So anything a model judges is shown as provisional until an objective check or you confirm it.
Demonstrated, verified
Demonstrated, provisional
Every marking keeps its rubric, its rationale and the name of the model that made it. A model's confidence is recorded and never turned into a mastery number.
- Verified at once
An objective check
Unit tests pass, a numeric answer matches, an answer key agrees
- Provisional
A model
A proof, an essay argument or an explanation is marked against a rubric
- Verifies or overturns
You
You read the rubric and the model answer, then confirm or dispute
Reviews that space out.
Once a concept is demonstrated, a review is scheduled. Pass it and the next one is further away.
- 7 days
- 16 days
- 35 days
- 75 days
After the fourth, each gap is 2.2 times the last, up to a year. A later release adds adaptive scheduling as a setting.
States go down, too.
Knowledge decays, and the record admits it.
| When | Then |
|---|---|
| A missed review on a retained concept | Back to demonstrated. A relearning review in 2 days. |
| A second missed review in a row | Back to practised. A practice session is queued. |
| A missed review on a demonstrated concept | Back to practised. |
| A practice attempt that falls short | No change. Any misconception is recorded. |
| A marking you overturn | The state is recomputed from the evidence that remains. |
One concept, twenty-six days.
The chain rule, on a calculus track, as the record would show it.
Day 1
Understood with help, provisional
Guided explanation. Two check questions answered, one hint used.
Marked by a model
Day 1
Practised
Three practice problems: pass, not yet, pass. One misconception recorded.
Numeric check
Day 2
Practised
Two problems aimed at the misconception, both passed. Misconception closed.
Numeric check
Day 3
Demonstrated
Two of three unseen problems passed, no hints. Review scheduled for day 10.
Numeric check
Day 10
Retained
Review with a fresh problem, no hints, passed. Next review on day 26.
Numeric check
Day 26
Demonstrated
Review missed. A relearning review is set for day 28.
Numeric check
What we claim, and what we do not.
Retrieval practice and spaced practice are among the better supported techniques for long-term retention in reviews of educational research. Rheticus is built around both.
We do not claim a measured learning gain from Rheticus itself. The app has not been released, and we have no data. The honest claim is about the system: it keeps the record, and it does not let familiarity pass for ability.
See what you have actually proved.
Free during the early beta, with everything unlocked. Join the beta and the Windows build comes to you first.