Look, I really don't like AI. I'm not sure anyone who is media literate enough to get that Cyberpunk is a satire could actually be a fan of how this stuff is being pushed. Most of its general public use is slop, and most of its non-general public use seems to ultimately harm the general public, economically or otherwise. Every now and then I test an LLM on a topic I'm really knowledgeable on, and none of the answers I've received have ever made me comfortable trusting it on a topic I'm not. I've yet to find a single are in my life where LLMs could actually make sense to use in a meaningful way.
I've also recently found I don't like the process of moding. I used to be able to work on a mod list for weeks, testing and tweaking. These days, I really just want to spend whatever time I can set aside for gaming playing the actual game. I know there are dedicated nexus/wabbajack modlists, but many of them are out of date, and the updated ones aren't exactly what I'm looking for.
Given the apparent recent development of LLMs like Claude in technical areas like coding, and seeing how modding is very much adjacent to that, I'm wondering if cutting down on the time it takes to research and put together a usable mod list could finally be an area where I could get some utility from these fucking things. I said that I've never received an impressive answer on a topic I know, but my deep knowledge isn't in any technical areas. Has anyone tried it with recent models? Did you get anything approaching a stable modlist? If I give Claude my PC specs and performance targets, point to an existing mod list like Welcome to Night City as a base, note some specific mods I'd like to incorporate/remove, and then give it further details on what I'm going for, am I going to be able to even remotely trust what it spits out? I'm not expecting it to be perfect, and I can troubleshoot minor individual issues, but is it going to be within the ballpark, or not even worth trying?