Even when it comes to non-reasoning LLMs such as Claude 3.5 Sonnet, there may be room for regular users to improve their prompting.