If it helps, the article “evolved” so I don’t really care that LLM’s had a part to play. I am setting up a development environment for Mezzano, the Common Lisp OS after getting it running on ARM64. I needed to understand the full CL toolchain to build an AI agent harness that could talk to Mezzano.

I figured out I could do this via SWANK. But kept hitting the same problem, the information about how all the pieces fit together is scattered across dozens of sources and nobody as far as I can tell had put a complete layered map in one place. Which I kind of already had from all the conversations and research I’ve been doing so I glommed it all together and posted it to r/lisp.

BTW the lisp community have been really helpful so I incorporated and continue to add all the corrections and pointers people have been giving. Case in point someone above pointed out vend which is an interesting approach that might be useful for my lisp harness project.

When I read an article, I am expecting to read the author's own experiences and insights they gained from them. Not the regurgitation of an industrial scale word generator.

> She still needed to know what questions to ask and prompts to give

Then publish the prompts. Let me enter them in an LLM of my choosing and see what bullshit it hallucinates and diff it against the 'article'.

> hopefully steered it right when it made up falsehoods.

"Hopefully"? Publishing something a stochastic parrot dreamed up under your name is ghost writing at best and spreading misinformation at worst.