Gavel & Gown: LLM Judge

​Your agent's output "looks good." Can you prove it? 🔨

​That's the gap this session closes. Installment 3 of the series is a laptop-open build for anyone shipping agent output they can't actually verify.

​Here's the problem: "looks good" is a vibe, not a QA process. So bad answers slip through, good ones get thrown out, and nobody can say why. The fix isn't a better prompt. It's a judge.

​So we build one. An unbiased LLM-as-Judge that turns "seems fine" into a measurable keep-or-revert verdict.

​(Your Judge will be judged by Judy the Judge. Kidding. Mostly.)

​You'll walk out knowing how to build:

• ​A deterministic process so the same input always gets the same ruling

• ​Evals that score outputs against a rubric instead of a gut feeling

• ​Loops that feed the verdict back in so the system sharpens itself

​This is the QA layer that separates agents you can trust from agents you just hope about.

​Bring your laptop. Leave with a judge.

​Best for: Founders, engineers, AI operators, and anyone building agents that need to be right, not just plausible.

​📍 Antler VC — 800 Brazos St #340, Austin, TX
🖥️ Austin in-person energy + recordings (free for members)
🎟️ Approval required — request your spot early

​Request to Join 👇