When I left my role at Mystery Science in the summer of 2021, I was left…a bit wanting. I expressly wanted to explore the EdTech space as my kids came of schooling age, but my role didn’t allow me to do this in earnest; Mystery was acquired by Discovery Education soon after I joined, and the acquiring team consolidated the teams within a year, with me left on the outside looking in. Having tasted but a nibble of the intersection between technology and education, I knew I wasn’t done with EdTech.
To satisfy my curiosity, I joined Kiddom, a series C startup focused on K–12 education, providing digital curriculum to schools as an alternative to traditional textbooks. As the VP of Engineering, my mandate was to scale and level up the engineering team as the company entered its next growth stage. It’s a familiar story; the processes and structures that worked to find that initial product-market fit are less well-suited for taking that product to more and bigger customers.
A bit of background first. Public schools are perpetually underfunded and underserved. Much of the money—and therefore attention—in EdTech goes to post-high-school ventures, where increasingly pricey college tuitions create space for alternatives, and everything from accreditation to tutoring to college applications is met with captive audiences. Despite the proven impact that quality education makes in children’s formative years, we don’t pay nearly the attention or spend quite enough during these K–12 grades.
Kiddom’s core insight is that the curriculum plays a massive role in kids’ educational outcomes. Our solution is to find high-quality curricula1, and then digitalize them into the Kiddom app to increase accessibility and interactivity, along with coupling the materials with relevant integrations like calculators and assessments. The holy grail, where learning can be personalized and customized automatically via technology—now with AI adding even more possibilities—comes up again and again in education circles.
There are plenty of apps that feature interactive learning that are attractive to kids, though many stretch their products into gamification2 to keep students entertained. A rare few, like Khan Academy, are built with much more structure and rigor, and work to support students as a supplemental resource to fill knowledge gaps. But the hardest, most complicated materials are the core curriculum: the primary set of subject-specific instructional content designated for each grade level from kindergarten through 12th grade, encompassing the entirety of the school year, adhering and aligning to educational standards set by governmental bodies3. It’s one thing to build a simple game to introduce the concept of addition to first graders, but quite a different level of undertaking to provide 180 days’ worth of lesson plans to ensure first graders grasp the 20+ defined mathematical abilities so they can build on these concepts next year.
Because of these strict requirements, there is a small, established set of publishers that provide textbooks to schools across the country with aligned curricula. It’s in some ways even worse than the college textbook racket because adherence to state-level standards makes for a high bar of entry, and school districts are a wholly captive audience. Add on politics and lobbying efforts, and the industry tilts heavily in the incumbent players’ favor.
It sets the classic story of entrenched giants versus digital upstarts. In practice, Kiddom has an app serving core curriculum, enhanced with components and modules that add interactivity and third-party integrations. The materials in question can be fairly dense and lengthy, encompassing both student-facing textbooks and teacher-facing instructions and guides. The team had to invent formats and data structures to represent this complex set of data, as there’s no real industry standard given how far it’s lagged in the digital adaptation of core curriculum. The reality of publishing higher-quality instructional materials is that it’s more complex, and we’ve been continuously improving our systems to account for this added complexity while optimizing for site performance.
To make matters worse, schools generally don’t deploy great WiFi or feature high-end computing devices; the target user has an underpowered Chromebook running over a flakey connection at sub-broadband speeds. Admittedly, it’s not that much worse than what mobile app developers have had to contend with for the past decade, so there are established patterns the Kiddom engineering team has been steadily applying to add resilience to the app. A positive side-effect to building for low-end machines, though, is that the app is more portable, particularly in less well-to-do school districts; teachers have commended our accessibility in libraries and even on tablets at home, at times when it’s physically impossible for students to lug heavy, ragged textbooks around after school.
Of course, I got the opportunity to play around with Generative AI. I’ve already sketched out my overarching thoughts on this iteration of AI, but I’m bullish on Kiddom’s approach to leverage LLMs. There are plenty of EdTech startups experimenting with incorporating AI in their products—usually some form of personalized tutoring, or student learning aid—but few that can train and fine-tune their models with quality curriculum data. With the aforementioned investments made in formatting jumbled messes of PDFs and XML documents into workable data, the team has been exploring content generation and augmentation on the curriculum side—naturally avoiding controversies around inappropriate student usage. To be clear, it’s still early days for Kiddom AI, but there’s a great team of technologists digging into the details of what LLMs can do in this domain and I’m pretty excited about the direction we’ve been heading.
So after the ~4 years I’ve now spent in the education space, I have a better grasp on the domain. Some of it reminds me of the intrinsic factors that constrain technology in healthcare: tech can make education more effective and more efficient, but there’s a resiliency in the classroom that sometimes isn’t met by EdTech apps and services. Instead of, say, software that can help write emails or automate accounting workflows, education software has to deal with the reality of many students managed by singular teachers, running on limited hardware, with users—both students and teachers, sometimes administrators too—who are not technically savvy and don’t react well to hiccups and don’t have time to incorporate new workflows.
But the field is also filled with people who truly care. These listed challenges would be impossibilities, if not for the dedication and passion of current and former educators who have moved to the industry. I’ve met teachers who, despite the learning curve of switching their lesson plans to digital, put in the work in their scant spare time because they felt like their students needed to move to the digital future. A number of our users spend inordinate amounts of time customizing and personalizing their curriculum, to fit the needs of their classrooms. Meanwhile, some of my engineers dedicated their careers to EdTech, and regularly put in effort to get the edge cases right because they’ve been in classes and know how disruptive an unexpected error would be. In direct contrast to many of the other industries that I’ve worked in, the mission and drive resonate so strongly with the people here, and it’s easy to get inspired.
For the third time in a row, just as the kids are wrapping up their summer vacation and prepping for the coming school year, I’m going on vacation with an indeterminate timeline. From what I can tell in the job markets, the tech layoffs that started in 2022 haven’t fully abated, which has shrunk the amount of opportunities outside of the niche around AI. That is to say—I have no plans right now, beyond tackling some of the long-overdue todos around the house. I do intend to do a lot more reading and blogging with my newly minted free time, so stay tuned for more developments there…eventually.
There’s currently a movement in education to transparently review and vet curricula; one prominent site is EdReports.org.↩
My kids are uniquely talented at finding these sites and asking for computer time to “learn math.”↩
The US complicates this further, since states can and sometimes do set state-specific standards that deviate from national ones just enough to necessitate a bunch of extra content.↩