Skip to content

Memory Match for a 5-year-old who can't read yet

A few months ago I sat down with my five-year-old to play Memory Match -- the classic game where you flip face-down cards and try to remember which ones you've seen. She loves it. She is also, at five, still a pre-reader. The pack we were testing had pictures on one side and the word for the picture on the other.

She paired the dog picture with the word DOG on her first attempt by recognizing the shape of "DOG" -- a short word with two round things and a tall thing. Not by reading. She matched the shape.

That was the moment I realized Memory Match was the right kid-targeted entry for the platform, and that it pulled the privacy-by-design rule into focus harder than any other game in the catalog. Both, in order.

Why this game first

The catalog has tactical wargames, social deduction, bluffing, and a card game that runs the AI personality lineup. Of all of those, Memory Match is the only one that an unsupervised five-year-old can actually play.

A few reasons:

  • The rules fit on a sticky note. Flip two. Match Take them. Don't match Flip back. Other person's turn.
  • The game self-paces. There are no clocks, no turn timers, no "you must act in twelve seconds." A kid can take thirty seconds picking which card to flip second and the game patiently waits.
  • The losing condition isn't punishing. You don't die. The grown-up has more pairs at the end. That's it.
  • The skill ceiling is real but invisible. Adults playing Memory against other adults is genuinely hard -- eight cards in flight is at the edge of working memory. The kid sees a fun matching game; the grown-up gets a cognitive workout. Same engine.

This is the kind of thing the platform is for. A board game for the family living room is shipping for a household of mixed ages and skill levels. Memory Match is the only game in the catalog where a five-year-old, an eight-year-old, and a forty-year-old can sit down at the same table and have a fair-feeling time together. Bluff requires lying skill. Werewolf requires arguing skill. Battleland requires reading a hex map. Memory requires nothing the kid doesn't already have.

Why the asymmetric pack matters

The default Memory deck has matching pairs: two cards with the same picture, or two cards with the same word.

The platform ships a different pack on top of the same engine: Words and Pictures. The cards are not symmetric. One side of every pair is the word DOG, the other side is the picture of a dog. CAT pairs with the cat. BUS pairs with the bus.

This is one line of code in the engine -- a display_value field per match_group_id, sharing one matching rule. It is also the difference between a game and a literacy companion. The five-year-old plays a Memory game. While she's playing, she stares at the word DOG and the picture of a dog and forms a connection between them that, six months from now, when she sees DOG in a book, she will remember was the funny matching game from when she was little.

I am not making a curriculum here. The product is not "early reading software." It is a Memory game that happens to have an asymmetric pack, which happens to teach proto-reading, which is a happy accident of the engine being general enough to support different display_values on either side of a match. The "happy accident" framing matters because it's what kept the design honest: the moment the game started saying things like "good job!" or tracking which words your kid had memorized, it would have stopped being a game and started being homework. It is a game.

The Hidden Flips mode

The same engine runs a second mode that turns Memory from a recall game into a deduction game.

In Hidden Flips, only the flipper sees the value of a card they flipped. The room sees the position pulse -- a square lights up in the flipper's color -- but does not see what was there. If the flipper finds a match, the pair is claimed publicly. If they don't, the cards flip back and only they know what was where.

This sounds small. It is actually a totally different game. You're not remembering what you saw; you're remembering what your opponent saw, and inferring from their behavior whether they found a card you'd want to know about. It is Memory played with hidden information, which means it is now a game adults will choose to play when there are no kids around.

Same engine. Same code path for a public flip and a hidden flip -- the only difference is whether the value gets attached to the public broadcast or only to the flipper's own view. The platform's role-filtered view rule (the server filters what each device sees based on the device's role) makes this trivial to implement. Bluff exercises the same rule on a much harder game with a public-claim layer and a server-only actual-cards layer; Memory Match exercises the same rule on the simplest possible thing.

The point of doing it on the simplest game is that the architectural beat is small enough to debug. If the role-filtered view rule has a bug -- if a Hidden flip ever leaks its value to anyone other than the flipper -- Memory Match will catch it before Bluff or Werewolf or the future games where the leak would be catastrophic.

The five-year-old has no account

This is the part that ties back to the post about device profiles, and the part that sharpened the rule when I was writing the kid-targeted game's UI.

The five-year-old is going to use the platform on a tablet her parent hands her. She is not going to type her email address. She is not going to remember a password. She is not going to "sign up." She is not going to upload a profile photo of herself. She has no idea what any of those things mean.

What she does is tap an avatar she likes (a dragon, a cat, a smiling cookie -- whatever the curated gallery has) and then play Memory. That's the entire onboarding. The tablet has a profile that belongs to the tablet, not to her. If a friend comes over and uses her tablet, the friend can change the screen name to her own. There is no concept of "the kid's account" because there is no concept of an account at all.

This sounds like a small thing. It is not. The default model for "platform a kid uses" is to require a parent-managed account, gate features behind parental approval, ship a settings page with a "kid mode" toggle, and call it safe. The settings get missed on fresh devices. The defaults are wrong. The kid finds the toggle and flips it. The teacher who has to onboard twenty-five tablets at the start of a school day doesn't have the bandwidth to verify each one.

The fix is not to ship the unsafe version with a gate. The fix is to not ship it. The kid-safe mode is the only mode. There is no settings page for it because there is nothing on the settings page that could turn it off. No chat to filter. No friends to approve. No profile to lock down. No demographic data to redact.

What the kid never sees

Working list of things missing from the kid's screen, deliberately and forever:

  • No "ask a parent to enable this." Nothing requires permission, because nothing dangerous is reachable.
  • No leaderboards. No global rankings. No "you are the 4,213rd best player." She is playing with her brother. The score that matters fits on the screen.
  • No friend list, no "add this player," no "send Grandma an invite to play next time." If Grandma is playing next time, a parent will text her a room code, like a normal human.
  • No chat. Not a moderated one, not a filtered one, not a sticker-only one. There is nothing to chat about -- she's in the same room as the people she's playing with.
  • No "you played for 47 minutes today, take a break" nag. The platform does not surveil her play time and report it to anyone. (If a parent wants screen-time limits, that's what the operating system on the tablet is for, and that's the right place because the OS already does it correctly.)
  • No demographic fields, anywhere. She has a screen name (which is "DRAGONLADY" right now, will be "PRINCESS" tomorrow), an avatar (the dragon, she's been on the dragon for two months), and a color (purple). That is the whole identity surface for a five-year-old on this platform.

Why this game proves the rule

Memory Match is the easiest possible argument for the privacy-by-design rule because the customer is a five-year-old. There is no demographic field on the planet that a five-year-old should fill in. Pronouns She doesn't know what those are. Age Her parent already knows. Location She's at home. Real name She is named "DRAGONLADY."

If the rule is right for the five-year-old, it's also right for the eight-year-old, the eleven-year-old, the fifteen-year-old, and the forty-year-old. The platform doesn't have a "kid version" and a "grown-up version." It has one version, and the one version is the one where there are no demographic fields, no chat, no friend lists, and no profile photos for anyone. The five-year-old gets the same product the forty-year-old gets, and that's the point.

Memory Match shipped the rule. Every later game inherits it.