Trivial parsers were trivial before LLMs, too.

Ironically, using the LLM to write trivial parsers loses out on the best part of the LLM: its ability to parse natural language into some sort of structure.

Using the LLM itself as the parser frontend seems more interesting all round...

"It's not just for _____, it's actually ____"

Cannot unsee this incredibly played-out LLM trope, especially in this case where it's used almost every other paragraph.

This itself was written by an LLM, no?