Parent view · trust-by-design

You see enough to support them. Not enough to crowd them.

Your child is learning with AI. This page shows what you get to see — and what they keep private — by their choice, not yours. Trust is the product, not surveillance.

No keystroke logs. The system never stores raw AI conversations for you. No final answers. You see what they're working on — not how they think.
M
Viewing as Parent · Mei's account
Child: Yuna · Grade 11 · 14 active threads · last share update 2 days ago
Visibility level: parent-shared only

What Yuna shared with you

3 of her 14 threads. The other 11 are private — she'll show you when she wants to.

"我每次想认真复习数学就刷手机停不下来 — 到底是我自制力差,还是手机本身就是设计来让人停不下来的?"
in progress
Started 3 days ago · 4 sessions · last activity yesterday evening
Visible to you

Question — what you see above.

What she's tried: read 2 articles, ran a 24h tracking experiment.

Current state: partially resolved. High confidence that app design has intent. Low confidence about what that means for her.

Next step she's chosen: 1-week deletion experiment.

She kept private

Specific unlock count from her experiment.

One personal reflection note she marked sensitive.

Her chat with AI about how the experiment made her feel.

(You can ask her about it. She doesn't have to share.)

Ask her about it · She'll see you opened the thread — no extra detail.
"Why does my school's history textbook say the Industrial Revolution started in 1760, but the Wikipedia article says it 'gradually emerged from 1760-1840'?"
resolved
Started last week · 2 sessions · she marked resolved 5 days ago
Visible to you

What she learned: Textbook dates are pedagogical anchors; historical events have fuzzy edges. Primary sources show industrialisation came in waves by sector, not on a fixed start date.

Confidence: high.

What changed her mind: reading two primary sources from the 1750s and 1820s.

She kept private

Her side argument with AI about whether "1760" was politically motivated.

(Optional context. The conclusion is what matters.)

Ask her about it · Maybe useful at dinner if she brings up history class.
"Should I take AP Computer Science next year, or AP Statistics? They conflict in my schedule."
paused
Started 10 days ago · 1 session · paused 9 days ago · note: "waiting to talk to dad"
Visible to you

What she's chewing on: a real scheduling decision. She paused on purpose to talk it through with a human.

What AI offered: three angles (career path, current strengths, what the conflict trades off).

What she wants from you: she explicitly noted "waiting to talk to dad". This is your cue.

She kept private

Her honest assessment of how good she is at each subject.

Why she's leaning one direction.

(She wants your input fresh — without anchoring it to what she already thinks.)

Open thread with her · The point of pause is to bring it to a human conversation.

Why you don't see more

The temptation is to show you everything — keystrokes, full conversations, time logs, attention scores. We don't. Each restriction is a design decision, not a missing feature.

01

You see questions and conclusions, not the conversation

The middle is where she experiments with bad ideas, says them out loud, gets corrected, and changes her mind. That's the learning. If you read it like a transcript, she stops being able to think out loud.

What this prevents — AI as adversarial homework log. AI becomes a private mind, not a reportable one.
02

She chooses visibility per thread, not per child

You don't get a global "show me everything" toggle. Each thread carries its own permission, set by her when she starts it. She can change it later. She can mark old threads private.

What this prevents — coercion by family setting. "You have to share everything or I'm uninstalling it" doesn't work because the system doesn't know how to comply.
03

You see paused for human tags loud and clear

When she explicitly tags a thread "waiting to talk to dad" or "ask mum about this", that's signal — that's where she wants you. The product surfaces these. AI does not try to be the human she needs.

What this prevents — AI dissolving the parent-child relationship. The cases where she actively wants you are not buried.
04

No keystroke logs. No usage time scoring.

The system does not tell you "she spent 47 minutes on this thread" or "she opened the app 12 times today". That's surveillance metric framing. Learning isn't a leaderboard.

What this prevents — turning learning into a productivity dashboard the parent monitors. The metric becomes the goal; the thinking becomes the cost.
05

You can ask. Not search.

You can open a thread, send "ask her about it" — she gets notified. You cannot search across her threads, scrape her writing, or export logs.

What this prevents — full-text search on a 16-year-old's inner life. The interface itself refuses certain shapes of curiosity.
06

The boundary moves as she grows

At 11, default visibility is higher. By 16, defaults move toward private-by-default. She can override either way. The age slope is encoded so trust expands without you having to fight for it.

What this prevents — surveillance habits hardening at the age when independence should be growing.
See how she learns → Why we built it this way