Every judge has been a student. Education is the lowest-explanation-cost domain in any hackathon.
At the Microsoft AI Classroom hackathon, all four top projects were learning tools. ChatEDU turns any document into tutoring, Insight converts lecture notes into quizzes and flashcards, Cuepal does the same live inside a classroom with GPT-4, and Dialogues Through Time makes it a conversation with historical figures. At the ADK hackathon, Brazil's Edu.AI won Latin America and Nexora-AI won EMEA. In 2026 Exam Dragon won the AI Agent Olympics.
Education wins for three reasons. Judges have all been learners, so the value needs no explanation. The output is easy to show: notes go in, a quiz comes out, and the demo is over in fifteen seconds. And ground truth is available — exams and textbooks are already structured, so accuracy can be demonstrated inside a short build window.
Takeaway: the field is crowded. When all four winners of an event are education projects, the domain alone differentiates nothing. Exam Dragon separated itself by naming a specific national curriculum in Estonia. Narrowing to which country, which grade, which exam is what creates distance. A general-purpose study assistant is the easiest thing to build and the easiest thing to lose in the pile.
Across five years of winners, polish and reliability beat flashy ideas.
Recent winners are built as teams of specialized agents, not single chatbots.
Winning projects solve problems someone is suffering from today.