Designing for constraints in education
A teacher is not going to beta test your app during third period.
I’m back! In the last month I rode a cable car for the first time, learned that I hate the wheel (pottery), and hot take: don’t understand why it’s harder to build a summer wardrobe compared to a winter one.
I also started a new job and launched a new workforce development cohort. These days so much of my time is centered in education, design, and AI. And I think, at least for the foreseeable future, I’ll lean into that a bit more. So many things are broken in this space, and I wonder if these free-form thoughts will help me, or someone else, make sense of it all and find better solutions.
Move fast and break things?
Tech is often synonymous with innovation, and innovation is often synonymous with big bets that break and remake. Innovation, at its core, is just doing something new. But somewhere along the way it’s turned into something that exclusively means scale, speed, and disruption.
But in education, the path of scale, speed, and disruption is challenging. When building for education you *have* to build with constraints so your product/service can functionally operate in the ecosystem that is The Classroom ™. Innovation in education isn’t about building fast. It’s about building with constraints, with care, and with context.
Here’s a non-exhaustive list of education constraints:
Stakeholders + users: Education, especially K-12, has more stakeholders than the average team would like to build for. You’re dealing with Students, teachers, admin, staff, and parents
Security: At any given moment there are a handful of federal, state, and county- level laws for child protection and data management. Your product has to be compliant with before even thinking about pitching it. No FERPA? FERgetaboutit. No COPPA? I’m gonna stoppa ya right there. (Thank you for reading that, it was probably painful)
Hardware: This looks like a shared cart of Chromebooks that has 90% of the internet blocked and is exclusively tied to school wifi. That is, if there are any left that haven’t been short-circuited.
Implementation: Who is running the professional development session to implement this? Will you ghost the teacher after the deal is closed? What does support look like?
Valid skepticism: Right now, educators aren’t that optimistic about AI, especially since the majority of students are using it to cheat. If you want to get buy in, your product marketing cannot be “-now reimagined with AI 🚀”
Designing for realism (which is not pessimism)
Designing with constraints doesn’t mean limiting your ambition. It means being realistic about the environments you serve. I don’t think the question is whether AI belongs in schools. It’s whether your AI can handle some of the hardest users and situations. We have to design for the standard days not the ideal ones.
Here are some half-baked rules I’m using as I build in this space. My work is primarily centered on adult learning, but these still apply:
Chaos mode: Can a substitute teacher use this? What if one student is with the nurse, another is 20-minutes late, one is on task, the other is not, and another has a note from home excusing them from using AI
Speaking of…The Parent Town Hall: Assume a vocal parent will bring a list of grievances to a school board meeting
Am I replacing a task or adding one? One of these is better!
Easy onboarding: Not to be dramatic, but if school staff can’t find value within five minutes it won’t make it past the front office
All speed and no fit makes Jack a bad product
With just $20 and a Red Bull you can create a surprisingly functional prototype in a weekend with tools like Replit and Cursor. This is great for exploring and prototyping, but shipping fast without considering your users doesn’t make your tool ready for a classroom. Vibe coding has shown us that code can work…without actually working. Especially when it comes to security and UX and quality code.
In education, speed to market means nothing if the tool can’t survive post-implementation. The cost of a failed implementation isn’t just money (which is already limited), or maybe a negative review in the App Store, it’s a wasted learning opportunity that honestly can’t be recovered and will likely require borrowed time to course correct.
To wrap
AI in education might be the most honest stress test for responsible tech/AI. You’ve got a diverse user base, high stakes, and zero tolerance for tools that don’t deliver. The psychology, behavior, and needs are still so new, but I’m noticing a product shift: more emphasis on 1:1 tutoring and supplemental out-of-classroom tools, with less focus on in-classroom edtech tools (compared to the ~2018–2022 wave).
And… that might just be the next post.
This week’s Study Notes:
A section for me to highlight what I’ve been reading, watching, and listening to. A media dump with an educational slant.
YCombinator released their Request for Startups for the Summer 2025 batch. AI for Education is listed and I pray that founders step into a classroom. Or maybe, even just read this post.
New York Magazine’s “Everyone Is Cheating Their Way Through College.” My favorite line, in relation to the amount of time it takes to write an essay with the support of ChatGPT: “I spend so much time on TikTok,” she said. “Hours and hours, until my eyes start hurting, which makes it hard to plan and do my schoolwork.” We love a screentime-conscious moment!
New Public is developing a new way to create hyper-specific communities, ones where scaling to billions of users isn’t the primary goal like the app mentioned in The Social Network etc., etc.
