It was against spec, but the only way to get the job done, and Fable said so in the handoff. I would have preferred being consulted first, but this was the second best option.
Squiggly emitter
Opus: A+(99%) I gave it Fable's demo code, and asked it for a 5 step plan to get it working in Fluoddity. No problem. 2 minor fixes were required in which I copy pasted a stack trace, and then described a bug. I'm deducting 1 (one) point for the fucking Y-axis inversion though.
I requested "these packages + sphere primitives" and the primitives weren't available, so Fable looked through NVIDIA's github for examples. It found a comparable demo called "spheres" and did what they did: implemented a non-RT core sphere intersector.
New Fluoddity renderer that uses the raytracing cores is running but guess what
I haven't ever been wowed by the results of web search before; I find opus to be a mediocre googler who mostly just pollutes the contex. But cuda/optix from python bindings was niche enough that it was really necessary here, and Fable killed it. Maybe I can finally trust a model with the internet!
I might even be able to speed up the path tracer significantly by querying the Optix acceleration structure we already have to build instead of splatting everything to dense voxels that take multiple gigs of VRAM and can take half a second to rebuild.
Here's Cortex in the new renderer.
Realtime demo of the RT cores based Optix renderer. Not quite 60 fps but it's way faster than the volumetric path tracer. I could probably also fold them together the same way that I do with sdfs and have both. Maybe all the teal particles only make clouds and everything else is beads, etc.
Dribbles
Oops! All Paperclips
Oops! All Paperclips
Oops! All Paperclips
Oops! All Paperclips
Fable: A-
Failed/Wasted a lot of my time on the first attempt w/o internet access. Oh well. Given web search and the results of my pip list, worked perfectly with one trivial fix from Opus. The key to its success was that it made an excellent compromise when it determined I asked for the impossible.
