And ollama were sketchy about not providing proper credit to llama.cpp, even though that’s all they are, a wrapper for it.

Ollama uses the llama.cpp backend for inference. I find Ollama noticably slower. Llama.cpp has had a built-in webui (used as llama-server) for a long time now so have owned the user experience too.

Yep now llama.cpp is copying Ollama