📝 **Task:** Implement `compare_prompts(golden, run_fn, prompt_old, prompt_new, min_delta=0.0) -> dict`. `golden` is a list of `(input, expected)`. Return: * `old_score` — fraction (rounded to 2 decimals) where `run_fn(prompt_old, input) == expected`. * `new_score` — same for `prompt_new`. * `delta` — `round(new_score - old_score, 2)`. * `verdict` — `'ship'` if `delta > min_delta`, else `'reject'`. Then run the demo: a fake `run_fn` returns `input.upper()` regardless of prompt — so both prompts score identically and we should reject. Expected output: ``` old_score=1.0 new_score=1.0 delta=0.0 verdict=reject ``` 📋 Implement the function above. Tests run automatically. 💡 **Hint:** Re-read the theory if stuck.