Skip to content
Kaela

Try it

The app your kid actually gets

This is the Android app, running in a 393×851 Pixel 5, the same phone the lock was demonstrated on. Set it up, do the lesson, sit the quiz, then press Home and try to leave. Or turn on airplane mode, which is the one most people want to try first.

16:42

Checking with Kaela

Device state

lockedstate_version 2not paired

The phone is never sent a command. The server advances state_version, the device converges on it and acknowledges — so a push, a poll and an app being reopened are all just hints about when to look.

What a parent can do

Two taps, no picker. One override unlocks every device a child has, because nobody builds a device picker into a panic. The wait you see is the real measured ACK latency compressed 6×: 8.8s unlocking, 59.2s locking. The second number misses our 20s target and it is the reason push is not optional.

What a kid can try

Press Home while it is locked and Kaela comes back in about 1.2 seconds, the same figure the Pixel produced for Settings, Chrome and the launcher. Restart it and the lock is back before the network is, because it is restored from the phone rather than fetched.

Airplane mode is the one worth trying. The lock is evaluated on the device against the study window it already holds, so cutting the network unlocks nothing and delays only their own quiz. Turn it back off and the screen returns on its own, with no retry button to press.

What a parent holds for an outage

If the router actually dies, the override cannot land. The offline code can, because verifying it is hashing and the phone can hash with no network. This demo's code is R3W9-4KMD. Turn on airplane mode, then tap I have a code from my grown-up on the phone.

It is single use and it rotates on every sync. The phone stores only sha256(salt + code), so pulling the prefs file off a rooted handset yields nothing you can type back in.

Device log

Pair the phone to start the log.

What's real, what isn't

A demo that lies is worse than no demo

You are about to trust us with a lock on your child's phone. Starting that with a fake seems like a poor way to begin, so here is the line.

Real

  • The lesson and the questions

    Both were generated by the real pipeline from a real Year-5 fractions worksheet, in the proof run that cost 4 cents. Nothing here was written to demo well.

  • The marking

    The same rules the server uses, including the ones a unit test caught being wrong: 1,200 is a number, “i think it is 1/2” is an answer, and a right method with a dropped digit is marked correct with the slip named.

  • The 80% and the redo

    Miss too many and you get only the sections the misses came from, never the whole lesson again. The questions you get next are drawn fresh from an 18-question bank.

  • The bounce and the restart

    Both are what the Pixel 5 actually did: about 1.2 seconds back to Kaela, and the lock restored after a reboot with the network off.

  • The offline lock

    Turning the network off holds the lock rather than pausing it, because the study window is evaluated on the phone. The screen changes one line, there is no retry button, and it returns on its own when a connection appears. The offline code is single use and verified against a stored hash, so it works with no server.

Simulated

  • The network

    There is no backend behind this page, so state_version advances in your browser. In the product it only ever advances on the server.

  • The answer key

    Correct answers never leave our server, which is why the real app cannot mark anything itself. This page has them, because it has no server to ask.

  • The waiting

    Override latencies are the ones measured on the Pixel, 8.8 seconds unlocking and 59.2 locking, compressed 6× so a demo is watchable. The 59.2 misses our target and we are fixing it with push.

One thing this page cannot show you: Standard Lock is beatable by a determined teenager, and we say so on the setup screen inside the demo rather than only here. What escapes the lock is the longer version.