This was my impression as well, the bit about GPU incompatibility with llama.cpp made me think he was in over his head.