Session details

Your cohort’s join link, date, and time appear here once the schedule is confirmed. After the session, the recording will be posted here too — along with office hours, if your cohort has them.

The session runs up to three hours, including a break.

Before you join

Everything you need is listed in Studio 3 Overview. Confirm it before we start rather than during.

How the session runs

  1. Opening questions — asked, not answered.
  2. A first exercise: identifying behaviors that can be verified independently. No code.
  3. Debrief.
  4. A second set of policies, where verification stops being straightforward.
  5. Debrief.
  6. A break.
  7. The Mirror Test — judging proposed tests together.
  8. Working with AI to shape the design for independent verification, still without writing code.
  9. Concepts, reflection, and the field experiment you take into your week.

The opening exercise

An online store calculates discounts for a cart. Discounts may come from customer status, coupons, products, quantities, promotions, and enterprise agreements.

Identify the behaviors that can be verified independently. Do not design classes, write code, or choose patterns. Your artifact is a list of behavior examples and what each one proves.

We will start with policies that are easy to reason about:

  • Preferred customers receive 10% off eligible products.
  • A coupon code SAVE20 gives 20% off eligible products.
  • Products marked excluded from discounts receive no discount.
  • Buying 10 or more of the same item gives 5% off that item.

Then they get harder.

During the session

One norm: we stay out of code as long as we can. If you notice yourself wanting to see the implementation, that impulse is worth observing rather than following — it is part of what this Studio is about.

Bring the same attention to your AI. When it offers code, ask it what behavior needs to be verifiable first.