Evaluation: Is It Actually Helping?

Write a 10-prompt test set with the expected behavior set down first, rerun it after every charter change, and keep a weekly scorecard using numbers you measured yourself.

What it covers

• A 10-prompt test set with the expected behavior written down first.
• Rerunning it after every charter change.
• A weekly five-question scorecard using numbers you measure yourself.

What you'll build

Scorecard.md, your test set run once and your first scorecard.