Yesterday I asked my locally running gpt4all "What model are you running on?"

Answer: "I'm running on Toyota Corolla"

Which was perhaps the funniest thing I heard that day.

>> print(“Hello, world!”.ai_reverse()) world, Hello!

First few versions of Swift kept changing how strings work because it's not entirely obvious what most people intend from the nth element of a string.

Used to be easy, when it was ASCII.

Reverse the bytes of utf-8 and it won't always be valid uft-8.

Reverse the code-points, and the Canadian flag gets replaced with the Ascension Island flag.

Character-level operations are difficult for LLMs. Because of tokenization they don't really "perceive" strings as a list of characters. There are LLMs that ingest bytes, but they are intended to process binary data.