If your GPT-5 results have felt a little… underwhelming, you’re not imagining things. Pietro Schirano, CEO of Magic Path, recently dropped a comprehensive GPT-5 prompting guide that explains why — and how to fix it. The short version? GPT-5 is way more sensitive to instruction style than earlier models, and most of us are still prompting like it’s GPT-4.

If you’ve been exploring AI tools like Claude, you’ve probably hit that wall where the outputs feel hit-or-miss. One day it’s genius, the next it’s a head-scratcher. That inconsistency usually points back to one thing: your prompts.