OctoFlow

OctoFlow is a general-purpose programming language designed with the GPU as the primary execution target, not a wrapper around CUDA or shader language. It runs on any Vulkan-compatible GPU (NVIDIA, AMD, Intel) in a single 4.5 MB binary with zero external dependencies. The language includes 766 standard library modules, 221 pre-compiled GPU kernels, and features like functions, structs, modules, and error handling.

The Loom Engine manages GPU memory across five regions with autonomous dispatch chains, keeping data in VRAM to avoid unnecessary CPU-GPU transfers. Code runs sandboxed by default with no file access or network unless explicitly granted. The maker notes the language was developed with AI assistance, though all architectural decisions remained human-directed.