StudyBuds is an AI practice platform for grades 5–10, built on a reversal: instead of answering questions, students teach an AI character that holds a misconception, and earn points for how deeply they explain it.

Adam Franklin came to us with the hypothesis and an early prototype — not a brief. We started with a one-week sprint: functionality matrix, information architecture, first concepts for the dual-panel practice environment. That turned into an eighteen-month engagement.

The hard part was never the chat. It was keeping the experience playful without losing the academic weight teachers needed to trust it. We built the practice environment around evidence — students pulled sources from a resource panel, screenshotted and annotated them, then made their case in chat. A milestone progress bar tracked how far a student got in the argument rather than how much they typed, with teacher-defined acceptance criteria sitting behind each step.

From there we designed the rest of the platform. Lesson creation and interaction types for teachers, reports with class insights and depth-of-knowledge rubrics, a public library with folders and an admin tool to curate it. Onboarding, character selection, leaderboards, and first-time tooltips for students. Plus the logo, character direction, landing page, and pitch decks.

In a Leanlab study with 600+ students, StudyBuds produced a statistically significant improvement in scores — some classrooms gained over 20 percentage points.