Former Google trio raises $5M to kill the 'wall of text' in classrooms

Generative AI was supposed to revolutionize education, but for most seven-year-olds, it just looks like a faster way to be bored by a paragraph of text. While Silicon Valley races to build bigger LLMs, a Zurich-based startup argues that the current crop of "tutors" are fundamentally designing for adults, not developing brains.

Sparkli, emerging from stealth today with a $5 million pre-seed round, is betting that the future of learning isn’t about a better chatbot, it’s about an engine that can hallucinate an educational video game in real-time.

Led by former Google product veterans Lax Poojary, Lucie Marchand, and Myn Kang, the company has already quietly deployed its "learning expeditions" into a pilot network of 100,000 students across 100+ schools. The thesis is simple but high-stakes: The $350 billion edtech market is saturated with digital worksheets, but it lacks a platform that can actually hold a child's attention without devolving into mindless entertainment.

The "Agency Gap"

The core problem Sparkli aims to solve is what they term the "Agency and Curiosity Gap." Standard AI tools like ChatGPT or Gemini are passive; they answer questions with text, which often kills a child's inquiry process instantly.

"Like most people, I turned to sources like ChatGPT or Gemini for answers, but I could see my son just switching off when confronted with lots of text - it just kills kids' curiosity," says CEO Lax Poojary, who previously worked on YouTube and Area 120 at Google.

Instead of a chat interface, Sparkli uses a multimodal generative engine. If a child asks, "How would I build a city on Mars?", the system doesn't write a list of bullet points. It generates an interactive "expedition", essentially a pop-up multimedia journey combining generated voice, visuals, and quizzes that lets the child make choices (e.g., "Do you want to build the oxygen dome first or the water recycler?").

The Business of Attention

This isn't just a UI tweak; it's a massive technical wager on latency and generation costs. Generating interactive media on the fly is significantly more expensive than streaming tokens of text. However, the pre-seed funding, led by Swiss VC firm Founderful, suggests investors are willing to subsidize the compute costs to capture the next generation of learners.

The investment is Founderful's first pure-play bet on edtech, signaling a shift in how European capital views the sector. While US giants like Duolingo and Khan Academy have integrated AI tutors, Sparkli is attempting to build an "AI-native" engine from the ground up, rather than bolting an API onto legacy curriculum.

The Risks and Returns

The stakes for Sparkli are tied to the broader skepticism regarding screen time. Parents are increasingly wary of "digital pacifiers." Sparkli’s counter-pitch is "active screen time", using the retention benefits of interactive learning, which research suggests can boost knowledge retention by 40-75% compared to passive consumption.

However, the path to monetization is steep. The edtech graveyard is full of startups that had great engagement but failed to convert schools or parents into paid subscribers. Sparkli’s strategy appears to be a hybrid: proving efficacy in schools (hence the 100,000 student pilot) to drive consumer demand when they launch the parent-facing subscription in mid-2026.

If the technology holds up, Sparkli could force a redesign of how we think about "personalized learning." If it fails, it will likely be because the cost of generating high-fidelity, safe, hallucination-free multimedia content at scale remains prohibitively high.

For now, the team is focused on the data. They aren't just trying to teach kids about Mars; they are trying to prove that an AI can be a better storyteller than a teacher with a textbook.