I’ve been ex­per­i­ment­ing with us­ing LLMs lo­cally for gen­er­at­ing datasets to test Harper against. I might write a blog post about the tech­nique (which I am grandiosely call­ing “LLM-assisted fuzzing”), but I’m go­ing to make you wait. I’ve writ­ten a lit­tle tool called ofc that lets you in­sert Ollama into your bash scripts. I think it’s pretty neat, since it (very eas­ily) lets you do some pretty cool things. For ex­am­ple, you can swap out the sys­tem prompt, so if you want to com­par...