At the start of a hackathon, a brief is only a sentence. By the end, it has become a room full of prototypes, half-slept conversations and ideas you would never have arrived at on your own.
That transformation was what stayed with me after HackLondon 2026. On 21 and 22 February, 230 students gathered at King’s College London’s Bush House for an overnight build. The public gallery eventually filled with 72 projects. Knowunity was there as a sponsor, and Gregor Weber and I were there to represent the company, speak with participants and judge the Knowunity track.
It was a weekend of firsts for me: my first time judging a hackathon and my first fireside chat. Being on the other side of the table changed the experience. Instead of trying to defend one idea, I had to understand many of them quickly—and work out which ones had found a real student problem beneath the features.
The brief was deliberately bigger than an AI wrapper
We asked teams to build a daily learning companion for university students: something that could help a student learn, stay on track and succeed, rather than simply placing a new interface in front of ChatGPT or Gemini.
That distinction matters. Information is abundant; sustained learning is still difficult. A useful companion has to understand more than the question in front of it. It has to help with motivation, attention, confidence, pacing and the social reality of studying. We judged the entries on the strength of the idea, the execution of the experience and the ambition of the technology.
The most interesting projects did not ask only, “How can AI explain this?” They asked, “What helps a student return tomorrow?”
Twenty-one teams chose the Knowunity track, making it one of the event’s most popular. What impressed me was not just the volume. It was how differently the teams interpreted “companion.” Some made it a tutor. Some made it a game. Some made it a promise to a friend. Together, the entries exposed a useful truth: learning support is as much a motivation problem as it is a content problem.
Four winning answers to the same question
The winners were not variations of one product. Each found a different reason a student might keep coming back.
First place
Epsilon Delta
A personalised university maths tutor built around an interactive whiteboard. Students could write or speak a problem, receive live guidance and generate a tailored visual explanation using Manim and ElevenLabs. The team designed the experience around different learning needs, including support for ADHD and dyslexia.
Its strength was coherence: the technical ambition served a clear learning experience. Epsilon Delta won a sponsored trip to Knowunity’s Berlin headquarters and dinner with the founders.
View Epsilon Delta on Devpost ↗Second place
StudyQuest
StudyQuest connected AI-generated lessons and quizzes to a complete game loop. Learning earned resources; resources upgraded an island base and a roster of companions; those companions then went on expeditions and fought bosses. It treated motivation as a design problem and built far beyond a points badge on top of a quiz.
View StudyQuest on Devpost ↗Second place
StudyArena
StudyArena borrowed the pace of social media and redirected it towards academic work: a vertical micro-learning feed, knowledge checkpoints, focus wagers and social study duels. A camera-assisted “proof of effort” flow checked the work at the end of a focus session. It was playful, provocative and unusually complete for a solo build.
StudyQuest and StudyArena each took a second-place prize of DJI Neo drones.
View StudyArena on Devpost ↗Third place
FriendUnity
FriendUnity started from the isolation of independent study. A live friend feed made other students’ effort visible, while a head-to-head Quiz Arena turned revision into a shared event. Its use of virtual body doubling was a reminder that a personal learning companion does not have to mean learning alone.
The team earned third place and £100 gift vouchers for each member.
View FriendUnity on Devpost ↗What judging changed for me
When you build, it is easy to confuse effort with clarity. You know every trade-off, every late-night fix and every feature that almost made it. A judge sees only the problem you chose, the experience you made and the story you can demonstrate.
The best projects made those three things reinforce one another. They could explain the student problem without a long preamble. Their most important interaction worked in the demo. Their technical choices made that interaction possible rather than existing only to make the stack sound impressive.
The fireside chat with Gregor gave us a chance to share Knowunity’s journey and what we are building, but the useful part was the conversation that followed. Participants asked thoughtful questions about the product, the company and the problems that remain open in education. Those questions continued at the sponsor table and, in several cases, showed up later in the projects themselves.
A sponsor gets something back, too
Sponsoring a track is often described in terms of what a company gives: the brief, mentors, prizes and access. HackLondon was a reminder that the exchange works in both directions.
In one weekend, we saw 21 small teams challenge our assumptions about what a learning companion could be. Their prototypes were early and imperfect—as every 24-hour prototype should be—but they made new behaviours tangible. Personalised video, social accountability, game economies and collaborative study stopped being abstract product ideas and became things we could click, question and discuss.
I left with a longer list of ideas than I arrived with, a new appreciation for the difficulty of judging, and a lot of respect for everyone who chose to spend a sleepless weekend building for students. Thank you to the HackLondon organisers, KCL Tech, the UCL societies, Hackathons UK, Major League Hacking, every mentor and sponsor, and especially every team that trusted us with a demo.
For my first hackathon as a judge, it set the bar very high.