Notes from building for schools.
We publish when we have something worth saying: AI that pays for itself in a school building, software that survives a school year, and what running a two-sided education marketplace has taught us.
The test we use before putting AI in a district workflow
Most school AI pilots come unstuck for the same unglamorous reason: everyone agrees to try it before anyone agrees what success would look like, so what gets measured is enthusiasm. Here is the test we run before writing any code.
Multi-tenant from day one, or never
Retrofitting tenancy into a live product is one of the most expensive mistakes in software. Here is why we made the call before the first feature, and what it cost us up front.
What families actually struggle with in ESA programs
Eligibility turns out to be the easy part. We asked Alabama homeschool families what wastes their time, and the answer reshaped the product roadmap.
Silent failures are the only kind that matter
An error nobody sees is a bug that ships and then lives in production for months. Here is a short argument for loud, unignorable failure.
Where we tell schools to leave AI alone
There are three kinds of school work where adding a model reliably makes things slower, riskier, or harder to trust. Here is what we suggest doing with each of them instead.