lmao Claude Code regex to detect people swearing at it.

[peter@thepit.social]

lmao Claude Code regex to detect people swearing at it. https://github.com/chatgptprojects/claude-code/blob/642c7f944bbe5f7e57c05d756ab7fa7c9c5035cc/src/utils/userPromptKeywords.ts#L8

[dataangler@vis.social]

@peter @dnorman Wait is an LLM just a giant regex / ifelse labyrinth?

[peter@thepit.social]

@DataAngler noooo no no, this is Claude Code, it's a wrapper for their LLMs

[susanleemburg@neuromatch.social]

@peter swearing at it uncreatively

[dnorman@cosocial.ca]

@DataAngler @peter what if it's just nested if/thens all the way down?

[thedansimonson@lingo.lol]

@DataAngler @peter @dnorman would be funny if true, but (to get pedantic) it would technically be a finite state transducer because it produces a new string based on a previous string. Basically a regex, but the state transitions produce an output.

Even so, regex are deterministic, while there’s stochasticity baked into LLMs.

Lots of other differences but that would be the most fundamental. The theoretical comparison could be fun!