6. Precision scales with model capability. A key insight from the thread: newer models like Opus expect more precise instructions and guess less than previous versions. If your prompts worked fine on older Claude but break on the latest version, your prompts were always vague.

→ View original post on X — @godofprompt