After reading through a bunch of LLM getting started tutorials and resources, here are a few observations:

• Simple code is fragile because the ecosystem is changing so fast (e.g. basic calls to a model on HuggingFace don’t work because of access changes)
• LLMs are unpredictable in unpredictable ways. LLMs are, by design, non-deterministic, but there are a host of other factors that seem to make LLMs behave in unexpected ways (e.g. other processes running while using a local install).
• Getting started resources tend to be confusing because the people writing them seem confused too. Am I an expert? No. But I can tell that others are struggling to make sense of what these things are doing.
• Implication and ethics in what people choose to use are altogether left out, save a few instances.
• Some is just substanceless. This “course” from Google feels like it was made by AI to game SEO (which Google runs so…)