Week 1 Session: Naming What You See

What is the difference between an observation and an assumption? The semester's first session works on a system small enough to read in one sitting, so the discipline — not the difficulty — is the point.

22 February 2027

Before the session#

Nothing to prepare beyond a working environment. The system used in this session is small enough to read in full during the session itself.

The incident#

You have been handed order-total, a single-file command-line utility, under 150 lines, that Northfield Mercantile’s back office uses to print the final amount due on an order from its line items, tax rate, and an optional discount code. It ships with a short README that accurately describes its inputs and output format. For one saved order — a discount-code order, not a tax-only one — the utility prints a total that does not match the figure on the order’s own printed receipt. The discrepancy is small: a matter of cents, not dollars. Nothing else about the order looks unusual.

What is known and what isn’t#

Known: the mismatch is reproducible on this one order, and does not appear when the same utility is run on tax-only orders. Unknown: everything about why. Several explanations are already imaginable — a rounding step in the wrong place, tax computed before or after the discount, a stale exchange rate, a copy-paste error in the discount arithmetic — and that range is the point. Guessing among them before recording what is actually in front of you is exactly the failure mode this session exists to prevent.

In the session#

Before you are permitted to open order-total’s source or propose a cause, you write a numbered log of what you can observe directly: the exact command run, the exact output, the figure on the receipt, and anything else checkable without inference. “The total is wrong” is not an entry; “the utility printed 84.70; the receipt shows 85.00” is.

The exercise is deliberately strict about ordering. An observation written after a cause has already been guessed at tends to smuggle the guess back in as fact — the session is designed to make that failure mode visible when it happens.

Afterwards#

The observation log from this session is the direct input to next week’s hypothesis work — bring it, don’t rewrite it from memory. You leave this session able to state what is observed to be true about a small, fully readable system, separably from what you suspect is happening.