Is this really a quirky limitation? LLMs seem to struggle with any sort of task requiring a "complete" list of anything. Which is a massive problem because those are super common tasks.