Paul Graham (YC, @paulg at the other place) described receiving an LLM-generated message as like getting the compiled code instead of the source code.

:)