AI in the classroom,
on the professor's terms.

Students were already using AI. Professors had almost no say in how. Most AI tools hand students finished answers and hand professors nothing at all, no sense of what the class understands, no way to set boundaries, no reason to trust any of it.

We watched that play out in courses at Middlebury and wanted the opposite: a tutor that refuses to do the thinking for you, and an intelligent dashboard of the class's knowledge that shows which topics students are struggling with while there's still time to act. It saves professors time instead of adding to their day. Professors set the rules: how much help the AI gives, and whether it can ever state a final answer. They students work inside them.

It's the first thing every professor and administrator asks, so here's the short version:

TA-EDU started at a MiddCORE talk by OpenAI that left Max frustrated: the ideas were big, but none of them looked like what actually happens inside a classroom. On a drive to Burlington, we asked each other what an AI tool would look like if it were designed around both students and professors. By the end of the drive we were partners.

What followed broke most of our assumptions. We talked to about 20 students and 20 professors. Students wanted help but were stretched thin by demanding schedules. They wanted to use AI and feared it at the same time, worried it would damage their critical thinking, yet certain they would fall behind without it. Professors worried about academic integrity, and about privacy. This fall we're piloting TA-EDU at a private high school, because the people actually in the classroom should shape and decide how AI enters it.