Better answers took longer - day 09
The better answer was also the slower answer. That is a product decision, not just a model decision.
Recent posts
The ābestā model lost - day 08
The model everyone calls ābestā lost my tiny test today. Learn why a cheaper, faster model might outperform state-of-the-art models for specific tasks.
How to test the ai llm judge - day 07
Evaluating the evaluator: why your LLM-as-a-judge is a system under test that needs its own validation.
I made AI Judge - Day 06
Evaluating the logic behind having an LLM judge another LLMās output. Establishing a helpfulness rubric and confronting definitions.
I tested the āboringā requirement - day 05
Putting predictable constraints and standard expectations to the test, evaluating how the LLM judge grades standard formatting and basic rules.