"Weed" is an AI/ML library in the style of vm6502q/qrack (now unitaryfoundation/qrack, on GitHub). I wrote the (C++) Qrack quantum computer simulator framework (now with +2.5M downloads of its ctypes Python wrapper) to have absolutely minimal dependencies and supply-chain vulnerability attack surface: it only requires pure language standard at bare minimum, with optional OpenCL or CUDA for (vendor-agnostic) hardware acceleration, and with optional Boost library inclusion for performance. "Weed" aims to provide the same standards and utility for AI/ML inference and back-propagation as Qrack does for quantum computing: never be locked into a hardware vendor, never be locked out of deploying on a platform due to lack of upstream dependency support, and let fastidious engineering and design point to the way to novel optimizations, all of which work together through a "transparent" interface of optimal default user settings.