← All sessions & projects
Project 4 · Due Sun 6 July · after Sessions 9 & 10

Project 4: ADR #3: Capstone Decision

The keystone ADR for your capstone: the single hardest decision, reasoned end to end, including failure and operability.

What this is

Your hardest call, fully defended.

Pick the one decision in your capstone you are least sure about, the one a reviewer will poke at first. Reason it the whole way through: the number behind it, the trade-off, how it fails, and how you would run it.

💡 In plain English

Out of every choice in your final design, pick the scariest one, the one you'd hate to be grilled on, and do enough work that you'd actually welcome the question instead of dreading it.

Do this, in order

The steps.

  1. 1
    Pick the keystone decisionThe one with the most riding on it, or the most uncertainty. If you're tempted to avoid it, that's the one.
  2. 2
    Frame it with a numberBring the capacity estimate that bounds the decision (your Project 1 skills). No number, no decision.
  3. 3
    Decide and justifyThe trade-off, named, with the axis it sits on. One committed sentence plus the "because".
  4. 4
    Add the failure storyWhat breaks, the blast radius, and whether it fails open or closed (Session 9). State it on purpose.
  5. 5
    Add operabilityOne SLI and SLO: how would you know in production that this is healthy? (Session 10.)

The ADR format

ADR, with teeth.

Title
ADR-003 ·
Context + number
what's at stake · the bounding estimate that frames it
Decision
"We will ______ because ______." (name the axis)
Failure
what breaks · blast radius · fails open / fails closed (and why)
Operability
SLI: ______ · SLO: ______
Consequences
accepted cost · open risk · revisit when: ___

How it's marked

What good looks like.

✓ Strong submission

  • The decision is backed by a real number.
  • The failure mode and blast radius are explicit.
  • There is at least one SLI/SLO.
  • You are honest about what you are still unsure of.

✕ Common mistakes

  • Picking an easy decision to dodge the scary one.
  • Happy path only, no failure section.
  • No way to tell if it is working in production.
  • Hand-waving the numbers.

Wrapping up

Hand in the keystone ADR.

This is the ADR a reviewer will read first in Week 6. Frame, decide, fail, operate. It pairs with the Project 5 design doc.

Logistics
One-page ADR + failure + operability · due Sun 6 July.
submit on Maven
Which decision?
The one you hope nobody asks about. That fear is the signal.
lean in
For the instructorHow to teach thisclick to open ▾

The talk-track for introducing and explaining this project to students, beat by beat.

  1. 1

    Raise the stakes deliberately: "The first two ADRs were practice. This one is the decision your Week 6 reviewer will attack first, so we reason it all the way down." Then immediately reassure: they already know the format.

  2. 2

    The why is about courage: "Pick the decision you're hoping nobody asks about. The discomfort is the signal that it's the one worth doing properly."

  3. 3

    Walk the steps; this ADR adds two new sections beyond P2/P3, failure and operability, so flag that explicitly. Tie them to Sessions 9 and 10 respectively.

  4. 4

    On the template, point at the Failure and Operability rows as the new, graded parts. "Fail-open-vs-closed and one SLO are what turn a student answer into a staff answer."

  5. 5

    Rubric: the avoidance failure is the big one. Name it: "If your capstone ADR covers an easy decision, you protected your ego and lost the marks." Reward honesty about open risk.

  6. 6

    Logistics: due July 6th, same day as the design doc, and note they reinforce each other, the ADR is the deepest decision inside the doc.