Replacing Python as the glue for direct to GPU ML tasks is nice, but what about DOOM 2 ways: - A C implementation (based on DoomGeneric) rendered into a GPU window. - A TypeScript port of DOOM that renders into the same kind of GPU window, using Electrobun’s WGPU APIs from Bun.
In a few days you no longer need Python, pip, venv, virtualenv, conda, mamba, poetry, pipenv, uv, pyenv-virtualenv, asdf, system python, homebrew python, conda python. Just native on-GPU ML/AI glue and workloads directly from Bun typescript. No webview, in Electrobun.