The Learning Lab reports what your practice shows and refuses to report what it does not. This is the part most people never read: where your own activity lives, what the five optional tasks measure, and why a chart sometimes says "not enough yet" instead of showing you a number.
Without a plan: The activity room and the evidence rules
How many FACETS you see follows your plan; the thresholds and the flags do not.
Watch: Your activity, and why a figure is withheld
2 min 7 s · presented by Jason, Mandarin Temple's AI tutor
Captions are part of the picture, under Jason. The full transcript is below.
Read the transcript
- Hi, I’m Jason, Mandarin Temple’s AI tutor. Let’s read a Learning Lab profile measure by measure, on a demonstration account.
- Recognising a word and producing it are separate skills. Here, recognising is twenty points ahead, and the Lab says what closes the gap: speaking and writing, not more reading.
- Your forgetting curve comes from flashcard reviews at real intervals: what you still recall after a day, a week and a month. It’s your own curve, not a textbook average.
- Every so often, lesson practice asks how sure you are before you answer. Set against the result, it shows whether your confidence can be trusted. Here, it tracks closely.
- Task engagement stability follows your accuracy through a sitting. Here it slips after about twenty minutes, so a short break there is likely to help more than pushing on.
- And your mistakes, grouped by what they have in common. The largest share here are words not yet familiar, then tones three and four: specific things to practise, not a general weakness.
- Select Activity to see when you actually study, hour by hour. This learner works most often around nine in the evening.
- The replay plays back every page you studied, in the order you did it, sitting by sitting.
- And the race to threshold: a facet crosses with eight answers and four in five of them right. Two have crossed so far; the other five have not yet.
- Jason can talk you through these figures. He sees the figures on this page and nothing else, and asking him a question uses AI credits.
- That’s your Lab, measure by measure: what your practice shows, and when you tend to practise. Open yours and see what it says.
Open the video guide on its own page, with its key moments
The activity room
Your own record, in order: what you practised, when, for how long, and what came out of it. Not a score — a ledger.
It is the room to open when a figure elsewhere in the Lab surprises you, because almost every surprising number turns out to be a coverage fact rather than an ability fact. A flat side on your profile usually means you have never been set that kind of work, and the activity list is where you can see that directly instead of inferring it.
Why it says "not enough yet"
Every figure in the Lab has a minimum amount of evidence behind it, and below that minimum it is withheld with the reason shown rather than estimated. A blank chart is indistinguishable from a broken one, so it never just goes blank.
The thresholds are different for each measure, which is the part worth understanding. A simple proportion settles down after a handful of attempts. A trend over time needs far more, because a slope fitted to five points is mostly noise. A confidence measure needs enough attempts in both the confident and the unsure band before it can say anything at all. One threshold for everything would be either uselessly cautious or quietly wrong.
Group figures need five people
Separate from the evidence rule, and for a different reason. A median over two or three learners is not a group statistic — it is one person’s number wearing a group’s label, and the smaller the group the more exactly it identifies them.
So below five learners, group figures wait. Individual rows and counts do not: a tutor can always open a learner they teach, which is the job. What the floor stops is a number presented as the cohort’s while being readable as an individual’s — the form that travels upward to an administrator who was never meant to see the individual.
The six things it will flag
Where there is enough evidence, the Lab raises specific, named findings rather than a general score. There are six kinds, and each one names a different problem with a different fix.
| Flag | What it means | What to do |
|---|---|---|
| Stopped practising | Twenty answers or more, then nothing in the last two weeks. | A conversation about the schedule rather than more material. |
| Confident but wrong | Fifteen or more answers rated “fairly sure” or “certain”, and fewer than 60% of them right. | The most valuable flag here. Slow down and check, because confidence is blocking review. |
| Not retaining | Twelve or more reviews three weeks or more after the last one, and under 60% recalled. | Regular review before new material. The schedule shortens the gaps after a miss, so keep up with it. |
| Recognises more than they can produce | Recognition at least 20 points ahead of production. | You can read it and cannot say or write it. Set writing and speaking work specifically. |
| Sudden drop | Accuracy 15 points or more below the previous 30 days, with ten answers or more in each. | Often circumstantial rather than ability. Worth a look rather than an intervention. |
| One dominant error type | One kind of mistake makes up 40% or more of ten classified errors. | The cheapest thing on this list to fix, because one explanation clears all of it. |
The five cognitive tasks
Five short exercises, a minute or two each, entirely optional and entirely separate from your learning: working memory, response inhibition, processing speed, sustained attention, and implicit sequence learning.
They are research. They are never inserted into a lesson, they do not change your schedule, your XP or your streak, and not doing them costs you nothing anywhere in the product.
They also need their own consent, which is asked for separately from everything else and can be withdrawn at any time. Where the learner is a guardian-managed child, that consent is the guardian’s to give rather than the child’s.
What the Lab is not
- Not a test — It reads the record your ordinary practice already produces. There is nothing extra to sit.
- Not a grade — It never changes your schedule, your XP or your streak.
- Not a ranking — No leaderboard and no position in a class. Comparisons are you against your own earlier window.
- Not reading what you wrote — It is built from outcomes — right, wrong, how long, how confident. No endpoint behind it returns free text to anybody.
Common questions
It is built from evidence and has not got enough yet. Each measure counts separately, so the fastest route to a full picture is variety across a fortnight rather than volume in one activity.