That's the crux with most discussions on here regarding LLMs. "I used gpt-4o with zero shot prompts and it failed terribly!" "I used Claude/o1/o3, I fed vario

That's the crux with most discussions on here regarding LLMs. "I used gpt-4o with zero shot prompts and it failed terribly!" "I used Claude/o1/o3, I fed various bits of information into the context and carefully guided the LLM" Those two approaches (there are many more) would lead to very different results, yet all we read here are comments giving their opinions on "LLMs", as if there is only one LLM and one way of working with them.