Friends Don't Let Friends Use Ollama
Summary
This opinion piece argues Ollama misrepresented llama.cpp, failed to credit open-source work, and pivoted to cloud with a closed-source app, leading to performance and compatibility issues. It provides a detailed critique of the Modelfile approach, registry bottlenecks, quantization limitations, and a cloud push, and ultimately promotes open-source alternatives like llama.cpp and related tooling for local LLM deployment.