An eight-agent knowledge engine that stops institutional knowledge from walking out the door — router, knowledge graph and dashboard, with 43 tests. A judge called it 'a product, not a hackathon project'.
The detail that stands out is the 43 tests. Most hackathon entries are built to survive a demo; a test suite means the team wanted to know when things broke. The judge's line about a product rather than a project reads like a reaction to exactly that. Eight agents behind a router feeding a knowledge graph also explains itself in one breath, because the division of labour is visible rather than implied.
Across five years of winners, polish and reliability beat flashy ideas.
Recent winners are built as teams of specialized agents, not single chatbots.
Accelerator seats, mentorship and exposure often outvalue the cash.
Slice the archive by year and the winning stack turns over visibly, generation by generation.
Summaries are written by this site. Project pages include demo videos, screenshots and the team's own write-up (videos may autoplay).