Can see the intent, but prompts like this betray a lack of understanding of how LLMs work at a fundamental level. You get wins from the behavioural influence towards specificity, patterns of thinking... absolutely. But it boils down to 'be smart, don't make mistakes', which doesn't boost a model's efficacy or make it perform like a better model. In fact, it spreads its attention mechanism over more things than the task you're giving it, which for long running or difficult things decreases its effectiveness.