GPT-3, announced in May 2020 by OpenAI, was a breakthrough in neural net modeling of natural language and natural-language-related tasks; the June 2020 API opened up GPT-3 use to outsiders, including myself. I extensively documented my experiences testing GPT-3 and learning how to use it primarily for creative fiction such as poetry; but I also tested some “nonfiction” uses (often in response to hyperbolic claims about what GPT-3 could never do). This page documents tasks like anagrams, queries based on premises described as ‘databases’, probing the problems with GPT-3’s commonsense and other tasks (often related to poor prompting, showing the importance of prompt programming, or the pernicious influence of BPEs)