The dev behind ollama is adamant that ollama doesn't use llama.cpp (based on a technicality -- it uses ggml, which is created by the same people behind llama.cpp and is the backend of llama.cpp)

He made such a big fuss about ollama implementing their own kernels and felt slighted about the online comments saying ollama didn't properly credit llama.cpp and it kind of left a bad taste in the mouth among the local inference community.

For me personally, it was this that made me avoid them at all costs: https://github.com/ollama/ollama/issues/11714#issuecomment-3...

Ok and now llama.cpp is copying Ollama

Innovation and competition are good things