listening is an active process of inference but often the output is silent token
like there's something subtly bad feeling about typing to a LLM chatbot and they only read it on send and then only to spew slop in response
like maybe it's my personality but i feel like ais push way too fast towards closure while also using way too many tokens to do so