I’m both surprised and not surprised. I’m surprised because these sort of problems with very clear prompts and fairly clear algorithmic requirements are exactly

I’m both surprised and not surprised. I’m surprised because these sort of problems with very clear prompts and fairly clear algorithmic requirements are exactly what I’d expect LLMs to perform best at. But I’m not surprised because I’ve seen them fail on many problems even with lots of prompt engineering and test cases.