I’m in that ballpark … 50+, and I have some experiments done with LLMs, too. Running a 2nd PC just for agentic workflows. Tried from LLM based auto complete to full vibe coding. But there are limits, especially in game development, as it’s one of the most complex software projects, so many dependencies. You have to post so many tokens, just to keep the LLM aligned. At some point it’s too much and context rot will appear, when there is too much information in the context window.
More and more I’m conflicted with this tech. All the social, economical and ecological impacts weight is so heavy on the scale with some indy game on the other side. I’m using it sparsely now.
Maybe it’s not the best idea to start another LLM thread here. IMHO, you don’t want this discussion in your HO-MA-SIM thread.
Please, read these threads, maybe it’s better to discuss it there.