How to Evaluate an AI Model
One AI model can win gold at the International Mathematical Olympiad while another leading model can only read an analog clock about half the time, so what do those record-breaking benchmark scores actually tell you? In this episode, host Emily Laird explains AI evals: how researchers test these systems, why leaderboards go stale within months, and why an AI agent saying “task complete” doesn’t prove the task got done. She also looks at the gap between feeling faster with AI and actually being faster. Then she gives you a simple way to run your own evaluation, so you can decide whether a tool…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.