Jackson is not the kid who picks up a book and disappears into a corner for an hour. When I ask him to read, he whines. When I push, it turns into a negotiation, "Why do I have to read 20 minutes, can't it be 10?" For a long time I told myself that was just who he is and he'd grow into it.
Then it started showing up on his reading tests, so I asked him about it. He walked me through his method. He doesn't read the passage. He goes straight to the questions, then scans back through the text for words that match and picks the answer that sounds closest. Admittedly, it's a smart strategy for an eight-year-old. It works right up until a question asks how a character felt, or why something happened, and the answer doesn't match the text word-for-word.
I knew more worksheets weren't going to fix that. If anything, they'd give him more practice at the shortcut. What I needed was for him to want to read, and to make reading the only way to win at something he cared about.
I told him he could design a video game
So I made him an offer. He would build a video game with Claude. And, he would set all the parameters and design elements. It just had to include reading. Everything else was up to him.
What he came up with is Monster Farm: Gate Quest. His monsters live on a farm next to a big barn with a giant square gate. There are 12 elements to collect, from Fire and Water to Shadow and Space, and every monster evolves into something bigger. Each gate is guarded by Stingo, a spear-carrying bee who tells terrible bee jokes before every battle. The final boss is the Glitch Beast Shadow King, who steals elements and glitches the world. And when you finally beat him, the credits roll, and three legendary dragons escape: a Candy Corn Dragon, an Ice Cream Dragon, and a Cake Dragon. I would never have thought of a dessert dragon. None of this makes any sense to me, but it resonates with him. And, better yet, it gets him excited.
Beyond all that every gate has a story written on it. The questions stay locked until he reads it and taps "I'm Ready!" If he wants to peek back at the story during the battle, it costs him Honey Drops, so reading carefully the first time is the cheaper move. Right answers power up his monster's attack. Wrong answers bring out Stingo's stinger, and many of the wrong answers copy exact words from the story. Word-matching, his old shortcut, walks him straight into the stinger.
The first night, I couldn't get him to stop playing. Then he started reading the stories out loud to his little brother, Austin. He has since told all of his friends about it, and a few of their parents have asked me to show them how we built it. That's part of why I'm writing this.
The prompt behind the game
This build worked differently from anything Jackson and I had done before. Our earlier projects, like his first video game, came out of fluid conversation. He'd throw out an idea, I'd type it in, we'd see what came back, and we'd go from there. This time, before we started, I asked AI to write the prompt for me. It was a new technique for me, and it changed how the whole session went.
The prompt had two layers. The layer Jackson saw was Claude acting as his game design partner. It asked him one question at a time, in kid-friendly language, with a few example choices plus "or make up your own." He had to read every question and supply every idea, so he was practicing reading before the game even existed. When Claude needed grown-up information, it asked "Question for Mom" and waited for me.
The layer he didn't see was mine. I told Claude which reading skills he needed to build, with inference at the top of the list. I described his shortcut in detail, so the game would be designed to make it stop working. And I made clear it should never feel like a test or schoolwork. The locked questions, the Honey Drops, and the copycat wrong answers all came from that hidden layer. The monsters, Stingo, and the dessert dragons all came from him.
That split is the Idea → AI → Play framework in action. Jackson was the creative director. I was the facilitator, bringing the context only a parent has. AI sat in the middle and built what we described. One more detail matters here: once it was built, the game runs on its own. Jackson isn't chatting with an AI while he plays. He's reading.
It also kept going after the first build. The game has a hidden grown-up corner that tracks his accuracy by skill. When his first round of gates turned out to be too easy, we brought that data back and built a harder set, with longer stories and clues spread across several sentences.
Did it work?
At his most recent teacher conference, his teacher told us his reading comprehension score had jumped 100 points. I'm not giving the game all the credit. He has a good teacher, he's getting older, and a lot of things are happening at once. But I really think it helped.
The change I notice most at home is smaller and, to me, bigger. He asks to play. A kid who used to negotiate his way out of reading now asks for ten more minutes of it, because the reading is how his monsters evolve and how Stingo goes down. He still isn't curling up in a corner with a book. He is reading carefully, on purpose, because he built a world where that's how you win.
All of this as schools and parent continue to debate AI
While we were building Monster Farm at the kitchen table, the debate about kids and AI kept going. Some school districts are blocking it. Others are rolling it out. Parents are asking hard, fair questions about safety, about cheating, and about whether kids will stop thinking for themselves. I don't think that debate gets settled anytime soon, and I don't think it should be rushed. Schools are still working out how and when AI belongs in a classroom, and most teachers haven't been given the training to use it well.
I understand that last part better than most, because of what I see at work. My day job includes building AI training for professionals at a large global firm. Big companies are spending real money teaching their people to use AI, and it's still hard. Smart, experienced adults struggle with the same things kids do: knowing what to ask for, giving enough context, and questioning what comes back instead of accepting it. If it takes that much support for adults with budgets behind them, it's no surprise schools need time.
But our kids can't wait for that debate to finish. AI is already in their world, and the train has left the station. Watching adults learn it every day has convinced me that kids need to learn AI the same way they need to learn to read: as a basic skill, built early, practiced often, with an adult beside them while they get the hang of it. If we wait until they're older, they'll figure it out alone.
Monster Farm happened to teach both at once. Jackson got better at reading by directing AI. Along the way, he learned what AI is good for, which was building the world he imagined. He also learned what it can't do for him, which is the reading.
How to try this with your kid
You don't need to be technical to build something like this. You need to know your kid, and you need to be willing to let them drive. Here's what made the difference for us.
Start with the problem, not the tool. Before you open AI, figure out exactly what's hard for your kid. "He doesn't like reading" was too vague to design around. "He skips the passage and word-matches the answers" gave the game something specific to beat. The sharper your description, the better the result.
Hand your kid the creative wheel. Don't show up with the game already designed. Give them the one rule that matters (for us, reading opens the gates) and let them decide everything else. The ownership is what keeps them coming back, so resist the urge to improve their ideas.
Ask AI to write the prompt before you start. Describe your kid, the skill you're working on, and the habit you want to break. Then ask for a prompt where AI interviews your kid as a design partner, while keeping your goals in a hidden parent-only layer. This was the new step for me, and it's the reason the session felt like play for Jackson and still did the job I needed.
Keep sessions short and let them complain. We started at about ten minutes, one gate per session, ending on a reward so he wanted more. When something was boring or too easy, that became the list of changes for the next round. His complaints are some of the best design feedback we've gotten.
The full step-by-step version, including how to set up the prompt, is on the Monster Farm activity page. You can also browse the rest of our activities for more ways to put your kid in the creative director's chair.
Jackson used to skip the reading to get to the questions. Now the questions don't open until he reads, and he's the one who built the gate.