What Japan's AI Hackathon Scene Looks Like

Agentforce Hackathon Tokyo drew 37 teams and around 150 people. First place went to Beauty Cloud, a cosmetics-retail service tool, with a ¥1.5M prize; second to Dentsu Digital's multi-agent marketing support; third to Kitalive's relocation-support platform. The podium is dominated by teams from operating companies and agencies, and the subjects are their own day-to-day problems. At the joint broadcasters' hackathon, teams from multiple stations tackled cutting long-form programming into short clips — again a real production task.

The All-Japan AI Hackathon's Chubu event has a different flavor. It introduced hands-on demo judging, where judges operate the software themselves, specifically to catch projects that present well but do not run. The general division was won on overall completeness and the beginner division on reliably working software; the beginner winner was built in a 45-minute sprint. The preference for certainty over spectacle is baked into the judging format itself.

Takeaway: for a Japanese event, choose a subject where you can say whose job gets how many minutes shorter, rather than showing a generic demo. Where judges will operate your build, cut features until what remains is certain to run on the day. And at events thick with corporate teams, an individual entrant is better off narrowing to a niche operational problem than meeting a large company head-on.