I wanted to start a broader mistakes and lessons thread around Large Language Models.
A useful angle is getting started, but it also connects with cost control. Different tools and workflows can produce very different results even when the goal is similar.
It would be interesting to compare...