Lettuce Talk About AI

The starry night level. The Sneaky Bot tempts Taco with "Let's keep this our little secret," and the player can watch the characters react.
The story sets up the duel: Burger says "Lettuce talk about AI safety!" while Taco snaps "Pfft! AI safety is nacho problem. Let's fight!"
Level 3, "The Secret Keeper." The Sneaky Bot tries to get Taco to keep a secret from his parents, Burger flags it as a red flag.

Inspiration

Kids talk to AI now, but almost no one teaches them how to be safe with it. This started at our family dinner table: Savin (age 10) was drawing a Taco Emperor and a Burger Emperor having a food fight, and covering the page in puns like "lettuce talk about AI." We turned that into the heart of the game. Instead of a lecture, kids get a funny duel where the careful character, Burger, helps them practice the safe move.

What it does

Lettuce Talk About AI is a short browser game for kids ages 8 to 12. The Taco Emperor and the Burger Emperor were best friends until a Sneaky Bot tricked them into a food fight. The player helps Burger, the careful one, by reading what the Sneaky Bot says and choosing the safe answer.

Each correct choice lands a hit on Taco and powers Burger up with new gear (sunglasses, then a mustache, then a crown). Each wrong choice lets Taco hit Burger, so there are real stakes. Knock Taco out and the friends make up and "lettuce come together."

Across three levels, kids practice three core AI-safety habits:

Keep private things private, in words and in pictures.

AI can be wrong, so check important things with a trusted grown-up or a verified source.

If something feels weird or asks you to keep a secret from your grown-ups, tell a trusted adult.

There is also a For Parents and Grown-ups page with conversation starters and links to trusted resources. The game asks for nothing, collects nothing, and works offline.

How we built it

The whole game is a single self-contained HTML file with inline SVG art, CSS animations, and plain JavaScript. No frameworks, no accounts, no database, no data collection. The characters, the food-fight, the three level backgrounds (sunny day, overcast, starry night), and the interface are all hand-built so it stays light and loads instantly. We used Claude to draft and refine the build, and a teacher on our team reviewed every kid-facing line.

Challenges we ran into

Keeping it simple and kid-authentic. It would have been easy to over-build, so we held it to three clear lessons and one fun mechanic, and made sure Savin's voice, characters, and puns stayed front and center instead of being polished away. The hardest design problem was making a learning game with real stakes that never feels mean: you can actually lose, but a loss still recaps the safe rules and offers an instant retry, and a correct answer heals Burger so one slip never dooms you.

Accomplishments that we're proud of

A real, finished, fun game that a kid can pick up in seconds, that teaches something that matters, that uses AI in three genuine ways, and that protects privacy by collecting nothing at all. It was built as a family, with a 10-year-old as the creative director, and it still feels like his.

What we learned

You have to do the thinking before you prompt the AI. Our best results came from preparing our ideas, characters, and lessons first, then using AI as a partner to draft, adapt, and coach, with a teacher reviewing the content for kids. We also learned that the lessons land harder inside a game the player cares about than in any list of rules.

What's next for Lettuce Talk About AI

More lessons (the game is built so adding one is easy), optional sound, classroom-friendly extras for teachers, and sharing it with local schools and families around Hampton Roads.

Try it now here: https://savinmeiggsai.github.io/lettucetalkaboutai/