> the more isolated, and the more textually representable, the better: system programming is particularly apt
I’ve written complete GUIs in 3D on the front end. This GUI was non traditional. It allows you to playback, pause speed up, slow down and rewind a gps track like a movie. There is real time color changing and drawing of the track as the playback occurs.
Using mapbox to do this straight would be to slow. I told the AI to optimize it by going straight into shader extensions for mapbox to optimize GPU code.
Make no mistake. LLMs are incredible for things that are non systems based that require interaction with 3D and GUIs.
Yep, they work especially if you instruct them to add into your program ways for them to "see" what it is happening. And the more embedding models are getting better, the better results we will get too, from their ability to "see". For now Gemini 3 is the best at this, but is not the best at coding as an agent, so we will have to wait a bit.