They literally vendored llama.cpp and they STILL called it "Ollama for *". Georgi cannot be vindicated hard enough.

didn't Ollama vendor Llama cpp too?

Most projects typically start with llama.cpp and then move away to proprietary kernels