The model was writing 108 blank answers per bill and I never looked
I tried typesafe.ai's decision model to make an LLM scoring step cheaper. It would have, by a third. The harness I built to test it found the other two-thirds.
#llm
In 2017 I wrapped a 1966 chatbot in a Promise and called it AI
I built a voice assistant on a Raspberry Pi in 2017 and put a 1966 chatbot where the brain goes. Ran the whole thing again this year next to a local model.