@feld @eevee WTF? If the user clicks the "explain" button, they would/should expect a correct explanation of what they see, but a LLM based "AI" will never satisfy this goal