Comment by astrange
20 hours ago
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.)
Here's an interesting experiment to try. Strip out the comments from your code, rename the variables to something generic, and feed it to Opus 4.5, Gemini 3, or GPT 5.2 with 'thinking' mode turned on. Ask it what the code does, and ask it to review it for possible areas of improvement.
If you don't have access to advanced reasoning models -- the kind that will crunch for 10 minutes before they give you any response at all -- put the sanitized code on pastebin and I'll try it with my accounts.
If you're right, then the models won't even recognize your code as a ray-tracer, much less be able to say anything meaningful about it.
Would have to strip a lot more than that, like every printf and function name.
https://github.com/mrvacbob/atrace
I see what you mean, I was assuming you were referring to a smaller code base.
I'm sure Paul Heckbert's business card is in-distribution, but it'd be a good example of what I'm talking about ( see variation at https://fabiensanglard.net/rayTracing_back_of_business_card/ ).