I think I found a good use case for an #LLM: I had an ancient export of recipes from #GourmetRecipeManager in HTML that I wanted to make more accessible using #Cooklang. I created an AI skill with cooklang specs (and a command to validate the output with cooklang #cli) and used it with Opencode to automate the conversion of the recipes. The results are pretty good (#MiniMax M2.5 Free), especially since the source material wasn't perfect. LLM output is acceptable when exactitude is not required.