SANbox

Logo

We set out to solve a problem we kept watching unfold in front of us. Homeschooling families, driven by genuine care for their children's education, were quietly overwhelmed. Parents lacked the time and expertise to guide every subject, and students were drifting; disengaged, screen-fatigued, and under-supported. The tools that existed were either built for traditional classrooms or simply replicated the same screen-heavy experience families were trying to escape.

The first difficulty we faced was conceptual. Every solution we designed defaulted to a screen. The average student's daily screen exposure already exceeds healthy thresholds. We committed early to a screen-light experience, and that single constraint shaped everything that followed.

The second challenge was making our AI behave like a teacher, not a search engine. A good teacher does not simply answer questions but they structure understanding. They pace the lesson, prompt reflection, and adapt in real time. We modeled our lesson flow around the principle that comprehension builds incrementally, where each stage of learning compounds on the last. Getting the AI to mirror that rhythm required us to think less like engineers and more like educators.

What we ultimately learned is that the hardest part of building something meaningful is not the technology but it is the humility to understand the people you are building it for. We learned that constraints produce clarity, that AI is only as thoughtful as the philosophy behind it, and that the products most worth building are the ones where someone genuinely needed them to exist.