The point of building for education

Improving education is the greatest lever we have to improve our civilization and society as a whole. Many of the smartest engineers and technologists of our generation are working hard to build AI that is far smarter and more capable than humans. Many more are working on B2B SaaS. If humans are going to have a place in an AGI world, they need to be educated, trained, prepared. This is the most urgent and meaningful problem of our generation.

Starting a start-up in the education space, though, has historically been a highly unprofitable thing to do, and the education investment landscape is a barren wasteland.

Of course, our belief is that this time is different. But let’s start by acknowledging why edtech has historically been difficult, and go in eyes-wide-open.

Why does edtech fail?

Students’ attention is impossible to wrangle. Parents are cautious and don’t want to experiment on their own kids. Schools take forever to acquire software and are some of the worst customers in the world to sell to. So much educational content is available for free, so the bar is high for a paid product. Parents, teachers, and students all have different incentives. Parents and teachers want evidence that the product works before committing. Kids churn every summer. Parents do not want their kids to use their screens as much. Parent acquisition costs are astronomical because the category is crowded and noisy.

We aren’t naive about this. We recognize the uphill battle ahead and the structural constraints – but we also believe that this technology wave will enable us to overcome them.

What’s different this time?

GenAI fundamentally changes everything about this equation. Multimodal, personalized, cross-platform AI tutors can feel more engaging than any edtech product ever. They can explain core concepts in infinite ways, adapt to the student’s understanding level and curriculum, and enforce mastery learning for true understanding. They can also build a real and lasting relationship with the student – far from the days of a cute cartoon character that you level up with XP points, this is the opportunity to provide a student their own Aristotle. They are also negligibly cheap to serve to users – orders of magnitude cheaper than human tutors (in addition to more reliable and easier to source).

Parents know that Gen AI is the future, and they want their kids to be a part of it, but they also want them to do it safely. A tutor built explicitly for safe, private, transparent use fills a huge gap in the market that ChatGPT, Claude, Gemini and Khanmigo don’t yet understand.

We’ve spoken to dozens of parents and tutored hundreds of students; the wedge is clear:

Disorganized but smart kids who do well on their tests but forget their homework, resist planners, and push back on their nagging parents. Executive function is a hidden crisis today – social media has driven up kids’ dopamine baselines and killed their ability to manage complex tasks and navigate uncertainty.

Parents are anxious about AI use. They don’t know what their kid is doing in ChatGPT and want a tutor that is transparent, safe, and aligned with school expectations. These pain points lit up in interviews and in ad tests.

Our hope is that if we start today with kids, all of those kids will grow up, and by 50 years from now, we will be the ones that have the market share with them. The only way we can get their market share is by getting parents to give us to their kids, which we can do because we are going to drive hard on the safe, private, human-in-the-loop AI. This is of course ridiculously speculative and pie-in-the-sky but we really believe it.