Part of the AI product hack day
A control room for the end of the hack day.
A day of building needs a good ending. I built one page to run it on the projector: who demos next, what got made and who won what.
The challenge
Everyone had built something and deserved a moment. I needed a fair way to choose who went next, and to make the results feel like a show rather than a spreadsheet.
My role
I designed and built it with Claude Code. The first version went in on the day of the event itself, while everyone else was building.
What changed
The wheel, the staged leaderboard reveal and the awards were a highlight of the day. Nobody looked closely at the AI scores, which is how it should be: they’re for fun.
What it does
The demo wheel
Spins to pick who demos next, with ticks, a fanfare and confetti. People come off the wheel once they’ve been up.
An AI judge
Reads each product brief and gives seven scores, a one-sentence comment and a “Most likely to…” title.
A staged reveal
The leaderboard starts hidden. The lower places appear from the bottom up, then the top five one click at a time.
After the day
Awards, a people’s choice vote and the feedback results. It stays online behind my own sign-in.
Keeping an AI judge safe
A product brief is someone else’s text, so the judge treats it as untrusted. It runs with no tools, in an empty folder, is told to ignore any instructions inside the brief, and has to answer in a fixed format.
The scores never decided anything that mattered. Recognition on the day was based on each person’s scope and starting point.