I want to have an evidence-based opinion about genAI and LLMs for coding in particular. Despite my misgivings, I want to find a way to do something significant with one, so I can go beyond trivial examples (both Claude and ChatGPT have shit the bed on every constrained problem I have given them). There are apparently lots of models I could try to run locally, offline. I live in a solar-powered house and I am not concerned about the energy impact I have seen from these. But one concern remains.