Half past eight on a Tuesday. A mother stands in the doorway of her son's room. Did you do any maths today? Yes. How is it going? Fine. The door closes.
She is not checking up on him. She really does not know what to do next. The report card shows a mark, the boy says "fine", and neither tells her what exactly he cannot do. We built the parent screen to answer that, and we have rebuilt it more times than any other screen in the product.
The difference that decides everything
"Maths is hard" and "I do not know how to help" sound alike, but they are different sentences. The first is a judgement of a child. The second is a request from an adult, and it can be answered.
Every version of this screen that we threw away had moved towards the first one. If you show an adult a little more than they need, the app stops helping and starts supervising. It happens unnoticed, one useful looking feature at a time.
What the screen says
Three things: whether there was a lesson today, what the child can now do without help, and what is worth going through together.
"Worth going through the eight times table together" is not a diagnosis. It tells the adult what to do tonight, and you can check whether it worked: in two days the screen either changes or it does not.
The screen also lists what the parent can see and what they cannot. The list is on the screen itself, not on a help page, and the child can open the same list. A boy who knows exactly what his parents can see behaves differently from one who suspects they can see everything.
What the screen does not show
What happened inside the lesson: which tasks came up, what he answered, how many times he made the same mistake.
What he said to the mentor. That conversation belongs to him, and so does his working. The parent sees only where he stopped, never how he got there.
The first question we get is always why the parent cannot see the answers. The honest reply is uncomfortable. A parent who can look will eventually look, and a child who knows that starts choosing the safe answer over the true one. Then the check stops measuring anything, and the whole route built on it goes wrong.
No numbers here either
Nowhere in the product does a number next to a child turn into a verdict as fast as here. We made the longer argument in a separate post. On this screen the effect is easy to describe: show a parent a number, and the evening conversation is about the number, why it fell this week and whose fault that is. The eight times table never comes up.
There are no rankings between children, for the same reason. The child's own screen does count the days practised in a row, because a child likes watching that number grow. But the count stays on the child's screen, and the parent never gets a line to ask questions about. A contest gives a short lift, and then a long slump for whoever finishes last. And the child who counts slowly already knows they count slowly.
An offer, not an inspection
The screen has one job: to make the evening start with something specific.
"Shall we look at the eight times table" can be accepted or turned down. "Did you study today" can only be endured. The mother in the doorway can say the first sentence because the screen gave it to her.
This is also why we send no notification saying the child has not studied for three days. The message goes to the parent, but it reaches the child, as a question through a closed door.
What stays on the child's side
He signs in with a code from his mother. He needs no email and no phone number, so there is nothing to lose.
He gets big buttons and four answer options. The wrong options are not random: they are the neighbouring cells of the same row, which is exactly where children make mistakes.
He can be wrong without getting a mark. The mentor says "good try, let's go through the part that went wrong", and gives him a hint to work from instead of the answer: eight sevens are eight fives and two more eights. And the app leaves him in peace: it never tells him he did not finish, and a missed evening is never held against him.
The same rule for a teacher and a manager
The family screen is the first case of a rule that applies across the whole product: the adult sees where a person is and whether they are moving, and the person keeps to themselves how they got there.
A teacher sees a class summary of the work they set, and nothing else: who has stopped coming, whose mistake keeps repeating, what to go over with everyone at once. The teacher does not see the courses a pupil takes at home or their talks with the mentor.
A manager sees which competencies are closed and how the assigned courses are going. There is no lesson content and no average score per employee.
It is one decision applied in three places. If the boundary were right only for children, it would be wrong for them too. Control gets in an adult's way just as much. Adults simply complain about it less.
Things we deleted from this screen
At first there was a feed of events: lesson started, lesson finished, answer wrong. On paper it was useful. On the screen it looked like a security camera, and we removed it completely.
Then there was a counter of lessons per week. It was harmless by itself, but within a day the question "why fewer this week" appeared next to it, and the evening was about a number again. What remains is "when the last lesson was", with nothing to compare it against.
There was even a "remind the child" button. It turned the parent into a notification system, and it excused us from making a product worth coming back to. We removed it.
Every deletion was harder than building the feature had been, because each feature looked useful. It was useful to the adult, and the child paid for it.
Where we might have it wrong
Some parents want more than we give, and we know it. In some families the adult knows the subject very well and could help better if they saw the actual tasks. Right now we are in their way.
The other gap: this screen assumes a child who studies alone. If a seven year old only sits down with a parent beside her, half of this separation means nothing, and we have not designed for that at all.
We drew the line in the child's favour. If it turns out that it blocks more help than it prevents harm, we will move it and say so openly, not with a quiet new checkbox in the settings.
