We Built an AI Judge for a Live Hackathon — Then Red-Teamed It

Arbiter watched 25 live demos at NEBULA:FOG:SINGULARITY, scored them with a three-model ensemble, delivered British-accented commentary to the room, and caught prompt injection attempts on stage. Here is the architecture, the five things that broke mid-event, and what we would do differently.

READ THE WRITE-UP
MORE ENTRIES DECLASSIFIED AS THEY CLEAR REVIEW