Waldridge did not initially expect GPT-3 and current LLMs to be as capable as they turned out to be; AI researchers who claim they were unsurprised are being misleading—the phenomenological power of LLMs genuinely exceeded expert expectations.