My big gripe with LLMs is the “high verbosity but low signal” style of their writing. Even the new 4.8 Opus writes like that, it’ll spit out a paragraph of verbiage for what could’ve been one sentence. I’m guessing… because we pay by output tokens? $-)