# Yuki's Session — Chapter 13 Worked Example

> The reroll timeline and diagnostic scorecard behind Yuki’s 1:47 a.m. dashboard-polish session in Chapter 13 of *The Defensible Decision*. Four iterations, five diagnostics, one shipped dashboard that runs below the WCAG contrast threshold. The scenario is about the CYCLE and the SCORECARD, not reconciled numbers.

**Wholly synthetic.** Any resemblance to a specific real late-night dashboard-polish session is coincidental; the pattern is drawn from typical practitioner-observed behaviour.

## Files

- [`yuki-session.csv`](yuki-session.csv) — flat rows: four reroll iterations + five diagnostics
- [`yuki-session.json`](yuki-session.json) — canonical scenario contract with source hash

## What it represents

A composite late-night session in which an analyst polishes a quarterly-review dashboard through four Copilot iterations without ever specifying a pass/fail criterion. The session ships in about four minutes. Chapter 13 uses the session to define <em>vibe charting</em> and to demonstrate the <code>5-of-5</code> pattern on the five diagnostics.

## Invariants (locked)

The scenario carries executable invariants that Chapter 13&rsquo;s release contract enforces:

- Exactly four reroll iterations.
- No iteration is verified against a rubric (every `verified` field is `false`).
- No iteration prompt contains a specific measurable criterion (no numbers, no named standards, no comparisons).
- Shipped contrast ratio is <code>3.1:1</code>, below the WCAG <code>4.5:1</code> minimum &mdash; the silent failure Yuki never surfaces.
- Acceptance rate is <code>1.0</code>; verification prompts across the session are zero; time-to-ship is under thirty seconds after the final render.
- Exactly five diagnostics are recorded, IDs `1..5` with no gaps.
- All five diagnostics fire (`fires: true`) &mdash; the vibe-charting verdict.
- Diagnostic names match the Chapter 13 vocabulary exactly: Ambiguity burden, Evidence chain, Telemetry trio, Repair loop maturity, Over-reliance signals.
- Every diagnostic carries non-empty evidence pulled from Chapter 13&rsquo;s <em>Running the diagnostics on Yuki&rsquo;s session</em> section.

## Why the invariants matter

Yuki&rsquo;s pedagogical value is the <em>5-of-5</em> pattern &mdash; the diagnostic scorecard where every diagnostic fires because the underlying failure has been visible from prompt one. If any prompt hardens into a criterion, if any iteration gets verified, if the shipped dashboard turns out to pass WCAG anyway, the scenario stops teaching what Chapter 13 says it teaches. The invariants lock the failure pattern so the figure and the chapter remain in lockstep.

## What it does NOT represent

- Actual dashboard data (the chapter&rsquo;s point is the session shape, not what the dashboard shows).
- A specific Copilot version&rsquo;s behaviour (the reroll pattern is medium-independent; the exact defaults vary with the surface release).
- Yuki&rsquo;s VP&rsquo;s later feedback (the chapter notes the review went fine, but the failure lands downstream when the wrong decision surfaces).

## Reproducibility

The dataset ships from the book source at `AlexBooks/applied-ai/ai-directed-viz/data/yuki-session.json` and is re-emitted here by `AlexBooks/applied-ai/ai-directed-viz/scripts/generate-yuki-session.mjs --site-root <site-root>`. The generator writes the source SHA-256 into `yuki-session.json` so drift between the book and this site is detectable.

## License

Released for teaching use alongside *The Defensible Decision*. Use freely.