Raising AI Kids: Issue 20

The Family AI Stack

The Table With Three AIs

David walked into the kitchen and saw three AI chats open at once: one on the family laptop, one tucked behind a homework tab on Sam's school device, and one on Sam's phone, face-up beside a half-finished bowl of cereal. He stood by the counter for a second, doing the quiet parent math of whether this was normal, impressive, or a small cybersecurity incident.

"I count Grok, the school account, and the local thing running on the family machine in the den," David said. "Should I be concerned or proud?"

Sam did not look up. "Different tools."

David waited, because that answer was both too short and probably true. "Different tools for different what?"

Sam shrugged. "Different things."

That was not enough of an explanation. But it was also the right instinct. Families are quickly moving from "Should my kid use AI?" to a messier question: which AI, in which account, for which kind of work, with which kind of information?

There is no longer one AI tool in the house. There are school accounts, parent accounts, phone apps, cloud chatbots, local models, coding assistants, image tools, and whatever your kid's friend sent them yesterday. Without a map, kids use whatever is open. That means no privacy strategy, no consistency, and no shared language for what changes when they switch tools.

David pulled out a chair and made it clear he was not starting a ban-things conversation. He wanted to know which tool belonged where. Sam suggested it was like folders. David said it was more like a kitchen. You do not keep cereal, steak knives, plates, and cleaning spray in the same drawer just because they all live in the same room.

Sam thought about that for a second and said, "So AI has drawers."

Exactly.

Why One AI Rule Is Not Enough

This is where the Family AI Stack comes in. Not a technical diagram. Not a command center. Just a shared map of where AI lives in your house and what each place is for.

That makes this issue different from the Family AI Operating System. The operating system is about house rules: what your family allows, what you check, what needs a parent, and what counts as honest use. The stack is about tool placement. It answers a more concrete question: when a kid is about to use AI, which drawer should they open?

That distinction matters because the same prompt can carry different risk depending on where it goes. "Help me study for a science quiz" is fine almost anywhere. "Summarize this screenshot of Mom's medical bill" is not. "Make my game character jump higher" belongs in the coding drawer. "Write my English essay for me" does not become okay just because the tool is local.

Kids do not need fear or hype. They need the working sense that AI is not one thing in one place. It is a set of tools with different tradeoffs.

The Drawers In The Stack

The cloud chatbot is the front drawer most families open first. Grok, ChatGPT, Gemini, or whatever your family uses sits here. It is powerful, convenient, and good for general questions, brainstorming, explanations, writing help, and ordinary learning. It is also where your prompts leave the house and go to a company server.

That does not make it bad. It makes it the wrong place for secrets. David told Sam to treat a cloud chatbot like email: useful for normal work, not where you casually dump passwords, medical details, private family conflict, financial information, or anything that would make you wince if it appeared in the wrong inbox. If your family needs the privacy baseline first, Issue 7 is the companion piece: Privacy & Data Hygiene for Families.

The school account belongs in its own drawer. If your child's school provides AI through a district account, keep that world separate. School work goes there. Personal family material does not. A school account may be monitored, logged, filtered, or governed by rules your family did not write. That is not a conspiracy. It is how school systems work, and kids should understand the difference before they mix school tools and family life without thinking.

The parent account is another drawer, and this one is easy to blur because families already share streaming logins and device passwords. But an adult AI account is not Netflix. It may have memory, payment access, work documents, private searches, connected tools, or adult context attached to it. If a kid uses it, a parent should be close enough to know what is happening. The rule is not "never touch my account." The rule is "this account carries adult context, so it gets adult supervision."

The local model is a different drawer because it runs at home, on your own machine. That can be useful for experiments, privacy-sensitive drafts, and helping kids understand that AI is software, not magic in the sky. But local does not automatically mean safe. A model running under your roof can still be weird, wrong, or missing guardrails. We covered that risk directly in The No-Guardrails Problem. The short version: know what model is installed, who made it, and why you trust it.

The coding sandbox deserves its own drawer because building with AI is different from chatting with AI. A coding assistant can generate files, change projects, run commands, and help a kid build something real. That is exciting, but it raises a different parenting question. Not only "What did the AI say?" but "What did the AI change?" If the tool can alter the thing your kid is building, then your kid needs to understand ownership, backups, and what it means to accept a change they do not actually understand.

The image and media drawer may be the one families underestimate. Image tools feel playful, which is part of the appeal. But they also touch identity, taste, realism, copyright, and what kids believe counts as evidence. A good stack does not need a long policy here. It needs a plain rule: creative experiments are one thing; fake proof, fake people, or private photos are another.

What Belongs Where

David did not turn this into a lecture. He opened a note and wrote the names of the tools Sam actually used. Then they sorted them together.

Grok went under general questions and random curiosity. The school account went under assignments and teacher-approved work. The local model went under private experiments and family drafts. The coding assistant went under building projects, not essay shortcuts. Image tools went under creative play, with a reminder that AI images are not evidence. The parent account went under ask first.

It was not perfect. That was fine. The value was in the conversation.

Sam pushed back on a few things. He said the local model felt slower, which was true. He said the school account was annoying, also true. He said Grok was better for "random stuff," which was not the most formal category David had ever heard, but it was honest enough to write down.

That honesty matters because a family AI stack is not about making a museum label for every app. It is about helping kids understand context. Where the tool runs matters. Who owns the account matters. What gets stored matters. What the tool is allowed to touch matters. What the school can see matters. Whether the tool can take action matters.

This is where the stack connects to The AI Receipt, but it is not the same habit. The receipt comes after AI touches the work: what did you ask, what changed, what stayed yours? The stack comes before AI touches the work: which tool should handle this at all?

Those two habits work together. Choose the right drawer first. Then keep the receipt.

The Stack Changes As Kids Grow

A younger kid may only use AI with a parent typing. A ten-year-old might have a shared chatbot and one safe coding project. A teenager may need school AI, cloud AI, local tools, image tools, and coding help, but they should be able to explain why they chose the tool they used.

That explanation is the whole point. AI literacy is not knowing every model name. It is knowing what kind of tool you are using, what it is allowed to touch, and why it belongs in that drawer.

The stack will also change as tools improve. A model that was too weak last year may become useful. A school account may gain better controls. A cloud chatbot may add memory settings you need to revisit. A local setup may become easier. The family rule should not be frozen forever. The map should be reviewed whenever the tools or the kid's independence changes.

When David finished the note, Sam took the laptop and added one line under Grok: random stuff, but not private stuff.

David laughed because it was not polished. But it was accurate enough.

Shared language beats vague anxiety every time.

What’s Next

Next issue, we are moving from tool choice to quality judgment: how kids learn taste in a world where AI can make almost anything look polished. The question is not just "Is this AI?" It is "Is this actually good?"

P.S. Sam left the stack note on the fridge. Under the local model, he wrote, "private-ish, but Dad still gets veto power." David thought about correcting it, then decided it was the kind of imperfect sentence a family could actually use.