3 ms·

So true... I get more mileage from just watching an agent work than building sophisticated LLM-as-judge workflows