By the end of this session, you can…
- LO 11.1Merge playtest findings and audit findings into one ranked backlog with impact × effort scored.
- LO 11.2Commit to a revision frontier — the top N items that fit within budget — and defend the cut line.
- LO 11.3Execute at least three revisions in the studio period; log the diff.
- LO 11.4Present your revised loop to a peer group in a 5-minute show-and-ask.
One table. Every finding. Ranked.
| Column | What belongs |
|---|---|
| ID | Short handle — PT-03, AU-02. Easy to reference in commits. |
| Source | Playtest / audit / instructor / own observation. |
| Finding | One sentence. No paragraph; those go in the source doc. |
| Impact | 1–5. How much does fixing this move a D2 objective forward? |
| Effort | 1–5. How many studio hours realistically? |
| Flag | Red if stopper (ethics, safety, content error). Non-negotiable top priority. |
| Decision | Do / defer / decline. Deferred and declined items have a one-sentence reason. |
A good backlog has items below the line. Items below the line have a reason they are there — low impact, high effort, out of scope, or a deliberate choice about what this version is not. A backlog with no declines is a to-do list, not a design.
Seven findings, one cut line, three declines
| ID | Src | Finding | I | E | Flag | Decision |
|---|---|---|---|---|---|---|
| AU-01 | Audit | Pager-interrupt uses red-only palette; fails color-contrast test for deuteranopia. | 5 | 1 | Stopper | Do · top of list |
| PT-03 | Playtest | Three of four target learners mis-read the escalation step as a routine pick. | 5 | 3 | — | Do · above line |
| PT-07 | Playtest | Two learners asked for the patient history to stay visible during the pager phase. | 4 | 2 | — | Do · above line |
| AU-04 | Audit | Feedback for a correct pick arrives one state later than for an incorrect pick. | 4 | 3 | — | Do · above line |
| PT-05 | Playtest | One learner said the scoring screen felt "punitive" after a second wrong pick. | 3 | 4 | — | Defer · needs a rewrite, not a tweak; re-run D4 before shipping. |
| OWN-02 | Self | Add a narrative branch where a senior nurse can be consulted. | 3 | 5 | — | Decline · high effort, uncertain objective lift. v2 candidate. |
| PT-09 | Playtest | One playtester wanted a leaderboard. | 1 | 3 | — | Decline · competing with the judgment objective; out of scope on purpose. |
Cut line sits after AU-04. Four items above, three below. PT-05 is a defer (real, but bigger than this session); OWN-02 and PT-09 are declines with reasons. The reasons are the design; the items themselves are almost secondary.
Score, sort, draw the line
| Time | What happens | Cue |
|---|---|---|
| 00:00–10:00 | Solo — transfer all findings into the backlog. One row per finding. | "No editing yet. Capture everything." |
| 10:00–20:00 | Solo — score impact and effort. Flag stoppers. | "If you rate >5 items at Impact 5, you have not really scored them." |
| 20:00–25:00 | Pair — swap backlogs; partner queries the top three rankings. | "Why is PT-03 impact 4, not 3?" |
| 25:00–30:00 | Draw the cut line. Above it: this session's work. Below: the decline/defer note. | "The line is a commitment, not a filter." |
Execute. Log the diff.
Work above the line, top down. After each revision, make a commit (or a dated entry in your change log) with the backlog ID and a one-sentence diff. The log is graded — a fast revision with no trace is worth less than a slow revision that is documented.
Today you will think of a great new feature. Write it below the line. You are not building it. This is the last revision week; what is not on the list is not in the game.
Three patterns break studio time: the backlog fight (a team re-argues rankings instead of building), the rabbit hole (the top item takes 90 minutes because the learner keeps expanding its scope), and the stopper freeze (a red-flag item paralyzes the team because they feel they cannot ship without fixing it). Cues in order: at 30 min, interrupt any team still on rankings — "execute the top two even if you'd reorder them later." At 60 min, flag any team on one item — "what is the 80% version?" At 90 min, name stopper freezes aloud; stoppers that cannot land in this session become explicit Known-Limits entries in D5, not the reason a team ships nothing.
5 minutes, 2 slides, 1 question
Demo the revised loop to your triad. Two slides max — before and after, or the revised state machine. Bring one question you want feedback on. Not "thoughts?" — a specific question. Your peers' job is to answer that question only.
Good questions / bad questions
- Bad
- "What do you think of my game?"
- Better
- "Does the revised pager-interrupt state visibly teach escalation, or does it feel like a chore?"
- Bad
- "Any feedback?"
- Better
- "Between the old and new feedback for incorrect picks, which lands the discrimination point faster?"
Before final presentations
The build-discipline handout for revision
Revision studio is when teams overbuild. Today's handout is the counterweight: structured implementation with Codex, task packets scoped tight enough to debug, and decisions about what to cut.
Codex-To-Three.js Build Playbook
The build-loop discipline for teams turning revision findings into bounded, testable code changes. Written to prevent the classic failure: asking Codex for a full game instead of a disciplined next slice.
Why this week Write your revision cycle as Codex task packets from this playbook. If the packet won't fit the template, the revision is too big.
Before you move on…
Four questions on this session's concepts. Choices lock on first click — formative only.