Great post, Rakesh! Just for fun, I asked one of your questions to ChatGPT (specifically "Compute 5^100"), and it got it wrong. However, when I responded simply with the word "wrong," it managed to correct itself. I think looking further into "LLMs correcting themselves" could provide some nice supplementary insights to what you've presented in this essay.
