G—P / FMX 430 / Semester / Week 14

Week 14 · Phase 4 — The Final

Production sprint II & playtesting

The last full studio session before exhibition. Sprint in the morning half, then the hard part: strangers experience your piece cold — in the headset and on the wall — while you watch silently and take notes. Then we triage what to act on, because one week is enough time to fix three things well.

Session Fri Dec 11 · 10:00 am – 1:50 pm · CCB-134 + VP Studio Bring your packaged build, your playtest script, thick skin
By the end of this session you can

2:00 · Brief Strangers, protocols, logistics

You can no longer see your own piece — you've been inside it for three weeks. Playtesting borrows other people's first impressions, but only if you run it as a protocol. Improvised playtesting produces improvised data.

Playtesting is a protocol

The rules: hand over the headset (or the wall) with a task, not a tour. Say nothing while they're inside — no hints, no "oh, that part's not done." Watch the first thirty seconds hardest; that's your exhibition guest's entire calibration window. Write down what they did, not what they said — behavior is data, opinions are weather. Ask your three scripted questions only at the end, and never argue with an answer.

Triage, or drowning politely

Every playtest produces more notes than one week can absorb. Sort ruthlessly: fix now — it breaks comprehension, comfort, or the build (crashes, sickness, "I didn't know I could move"); exhibition polish — it would make the piece better and costs under an hour; next project — it's real, it's right, and it's out of scope. Bucket three is not failure. It's the most professional list you'll write this semester.

Exhibition is logistics

Week 15 runs on boring lists: builds installed and verified cold on the actual hardware, headsets charged, guardian boundaries set, a reset procedure between guests, signage, and a fallback if hardware dies mid-show. A brilliant piece that won't launch in front of guests exhibits as nothing. Packaging always surfaces bugs the editor hides — which is why it happens today, not finals morning.

Thanksgiving contingency

In years when the holiday lands on our meeting day this week, this session runs async: sprint at home on the same contract, post your packaged build and a 60-second capture to the class channel by Friday night, and run two remote playtests (roommates count — they're ideal strangers). The AI lab folds into your sprint as usual; playtests on the wall shift to install day in Week 15.

State of the Art · The sectors nobody told you about

Entertainment is the visible slice; healthcare simulation, manufacturing digital twins, and smart-city planning hire spatial computing graduates too — job-board estimates put median spatial-computing salaries around $141k (ZipRecruiter; treat as indicative). Discussion: describe your final project in the language of three different industries.

ZipRecruiter data · Spatial computing 2026 guide

2:45 · Sprint Package first, then build

Inverted priority today: get the packaged build installed and launching before you touch features, because playtests start at 4:00 whether you're ready or not. The agent earns its keep on packaging failures — build logs are long, boring, and exactly its kind of reading.

  1. Package for the exhibition target

    Package the project for your venue — Quest/Vive build for headset pieces, the wall's project format for VP pieces. Do this first; packaging always finds something.

  2. Diagnose failures with the agent

    If the build fails, paste or point Claude Code at the packaging log: "find the actual error in this log and propose the smallest fix." Verify by re-running the package — a build that completes is the only acceptable proof, not the agent's confidence.

  3. Install and cold-launch

    Install on the actual exhibition hardware and launch it cold, from off. Note launch time — your guests will stand through it. If it only runs from the editor, it doesn't run.

  4. Define the reset procedure

    Script how the piece returns to its starting state between playtesters — a menu, a level reload, a relaunch. If reset requires you and a keyboard, fix that now; you'll perform it forty times at exhibition. An agent can wire a simple reset input if you spec it precisely.

  5. Stage your playtest station

    Print or pull up your script: the task prompt you'll give testers, your three end questions, and a blank observation sheet per tester. Decide where you'll stand — able to observe, unable to hover.

  6. Sprint the remaining hour

    Whatever time is left goes to your sprint board — highest-risk item first, the Week 13 contract still in force: commit before agent sessions, review every diff, demo path over polish.

  7. Capture documentation of the true state

    Render this week's MRQ pass and grab headset or wall footage of the packaged build. Your documentation edit needs the real thing, and next week there's no time to shoot it.

4:00 · Playtests Strangers in the work

Rotation in pairs: everyone tests two pieces and hosts their own. Headset pieces run at stations; VP pieces run on the wall with the tester framing shots through the camera monitor. Hosts follow the protocol — task, silence, observation notes, three questions at the end. Testers owe honesty: what confused you, what you wanted to do that the piece refused, and the moment you'd walk away at an exhibition. Notes go straight into the repo; memory is not a capture format.

5:15 · Crit / share Triage together

Each person reads their rawest playtest observation aloud and sorts their notes into the three buckets on the board — fix now, exhibition polish, next project — with the room allowed to challenge exactly one placement per person. You leave tonight with your final week's to-do list already written, and it should be shorter than you fear.

Homework — due before Week 15

TaskDeliverable
Work the "fix now" bucket, then lock the build — no new features after the fixes land.Final build installed and cold-launch verified on exhibition hardware.
Cut the documentation edit from your MRQ renders and captures.2–3 minute portfolio-ready edit, exported and linked in the repo.
Assemble the paper trail: process book plus the full-semester AI log with its one-page reflection — where the agents made you faster, and where they made you worse.Process book + AILOG.md with reflection, committed and complete.

Resources