> It has nothing to do with "raytracers are well-represented in the training set" though. I find it so strange when people get overly specific in an attempt to sound savvy. You should be able to easily think of like five other ways it could work.

Can you elaborate? Your first sentence seems to be saying that it's basically irrelevant whether they have been trained on text and code related to raytracing, and I have no idea why that would be true.

I didn't say "text and code related to raytracing" though. I (and the parent post) said "raytracers".

It's more important whether it knows basic concepts about computer graphics, linear algebra, etc. Reading the code of a raytracer is not that helpful because it's hard to extract general concepts from low level code like that.

Besides that, it has web search and research tools.

I just fed Claude Opus 4.5 the source of a raytracer I wrote actually, and it had reasonably good comments on it, but it knew less than I know and its updated version had a few more bugs and was missing non-obvious optimizations I'd added. (In particular it loves writing FP math as all doubles for no reason.)