Tekku is raising.
AI-native coding for kids 8 to 14. Kids plan, build, ship, and learn on the same surface. The AI shows its thinking. Parents see what their kid actually learned. Every number on this page resolves to a cell in the live model.
Kids are already using AI. They are using it badly. Claude and ChatGPT write code an 8-year-old cannot read, run, or fix. When it breaks, the kid quits. Nothing is built, nothing is learned, nothing is shipped. Every parent we have spoken to describes the same scene. The screen-time debate at the dinner table gets louder every month.
The category as it stands is split between scripted code-lesson apps that teach syntax without shipping anything, and chat tools that are not safe, not pedagogical, and not measurable by anyone who loves the kid. Tekku is engineered for the middle: a glass-box AI that plans in the open, writes real code, and lands every session with a deployed URL the kid can text to a grandparent.
The compound sits one layer below the product. Every session labels what a kid actually learned. That concept data becomes the weekly parent report, the school-standards alignment, and eventually the only dataset in the world that tells you how kids learn to build with AI. The product is the wedge. The data is the moat.
Tekku's flywheel
What the pilot shows
Market opportunity
Kid creative software plus AI-education has a long tail of well-funded incumbents and no direct competitor in the kid-safe AI-native build surface. The gap is not small.
Competitive landscape
Scratch, Tynker, code.org, and the adult-coding assistants that kids get handed. Each lane leaves the middle empty. Our teardown walks through why.
Economics
The Builder plan at $149 per month is the wedge. Contribution margin improves with every Anthropic and OpenAI price cut. The sensitivity to a bad churn month is what we watch.
Moat
Concept-data flywheel, parent trust compounded through weekly artifacts, and a safety posture that is expensive to retrofit. The gallery becomes distribution.
Safety and COPPA
FTC COPPA modernization lands April 22, 2026 with $51,744 per violation per day. Every kid-AI product built before the deadline has retrofit debt. Tekku starts from the right priors.
Team
One founder inside the problem, a Quinn-mode agent team shipping real work, and a hiring plan that converts the agent velocity into a human eng and trust team without losing speed.
Expansion
Builder plan funds the founding cohort. Workshop tier opens the schools lane. Adjacent products follow the concept graph. The stage ladder is gated, not assumed.
At Year 5, base case, Tekku is a premium home product with a reliable school motion on top. Parents pay $149 a month without thinking about it because the weekly email still shows them a thing their kid made and a concept their kid learned. The Workshop tier runs in a mid-three-digit count of classrooms with curriculum alignment that does not require a teacher to become a prompt engineer.
The concept graph has compounded into the only dataset in the world that describes how children learn to build with AI. That dataset funds a second product we have not named yet and a defensible position against every larger incumbent that decides, too late, that the kid lane was real. The seed is the vote that makes the base case the reasonable case. Revenue at Year 5, base scenario: $141.4M ARR, 79,596 paying subscribers, one founder who still uses the product with his kid every Sunday.