G—P / FMX 430 / Semester / Week 01

Week 01 · Phase 1 — World Building

Orientation: the spatial computing landscape

Where the field actually is in 2026, what we're building this semester, and an install party that leaves every machine in the room ready to make worlds — with an AI agent wired into the toolchain from day one.

Session Fri Sep 04 · 10:00 am – 1:50 pm · CCB-134 Bring your laptop, your Epic account password, patience for downloads
By the end of this session you can

2:00 · Brief The landscape, honestly

"Spatial computing" is a marketing phrase and a real thing at the same time. The real thing: software whose interface is space — where position, scale, and your body are the input devices. This course approaches it as a studio production practice, not a survey.

The map we'll use all semester

One engine spans all four: Unreal. That's why this is an Unreal course. UE 5.8 is the final release of the UE5 generation, and it ships with something no engine has had before: an official way for AI coding agents to operate the editor (the Unreal MCP server). You are the first section of this course to have that.

State of the Art · The 2026 hardware paradox

Headsets are shrinking while glasses explode: standalone headset shipments are in a fifth straight year of decline (Omdia forecasts −15% in 2026), yet AR smart glasses grew 136% year-over-year and Meta and Google have pivoted toward glasses. Discussion: you're building for Quest — a shrinking but still 58%-dominant install base. Why is that still the right training bet?

Omdia, June 2026 · Counterpoint tracker

What "AI in the toolchain" means here — and doesn't

You will use Claude Code and Codex CLI every week, in a structured lab block. You will also take a practical exam in Week 7 with the agents switched off, because a director who can't do the work can't direct. The deal, in one sentence each:

Read the full policy on the AI Lab tab tonight — it's part of this week's homework and it is graded material all semester.

2:45 · Guided lab Install party

Downloads are big; we start them early and layer the small installs while they run. Follow the order below — it's arranged so nothing blocks anything else.

  1. Epic Games account + launcher

    Create or sign in at epicgames.com, install the Epic Games Launcher, and start the Unreal Engine 5.8 download first — it's the largest (plan for 40+ GB with debug symbols off). Uncheck target platforms you don't need; add Android if you plan to deploy to Quest from your own machine (we'll revisit in Week 10).

  2. Twinmotion

    From the same launcher, install the current Twinmotion release. Education entitlement is free — activate it with your UT email.

  3. Git + GitHub

    Install Git (macOS: xcode-select --install; Windows: Git for Windows). Create a GitHub account if you don't have one, and set up authentication with the GitHub CLI: gh auth login. Every project in this course lives in version control from the day it's created.

  4. Claude Code

    Install Node.js LTS, then:

    npm install -g @anthropic-ai/claude-code

    Sign in with the account arrangement discussed in class. Verify with claude --version.

  5. Codex CLI

    Install per the current OpenAI docs and sign in. We use two agents deliberately — comparing them is a graded exercise in Week 9.

  6. Create your first project

    When the 5.8 download lands: Launcher → Unreal Engine → Launch → Games → Blank, Blueprint, no starter content, name it FMX430_Sandbox. Let it compile shaders — this takes a while on first run; that's normal.

  7. Put the project under version control

    In a terminal, from the project folder:

    git init
    curl -o .gitignore https://raw.githubusercontent.com/github/gitignore/main/UnrealEngine.gitignore
    git add -A && git commit -m "Fresh UE 5.8 project"
    gh repo create fmx430-sandbox --private --source . --push

    The .gitignore matters: UE projects generate gigabytes of rebuildable files (Binaries, DerivedDataCache, Intermediate, Saved) that must never be committed.

If your machine can't run it

Don't burn the session fighting hardware. The CCB-134 lab machines have everything installed; pair with a neighbor today and see me about lab access before you leave. The course is designed so lab machines alone are enough.

4:00 · AI lab First contact

This week the agent only reads — no edits, no editor control yet (that's Week 2). The goal is to learn what it's like to interrogate a codebase you don't understand, which is exactly what a fresh UE project is.

Prompt 1 · Project anatomy
Look at this Unreal Engine project. Explain what each top-level folder and file is for, which ones are generated and safe to delete, and which ones hold my actual work.

Run claude from inside FMX430_Sandbox first. Notice that it reads files before answering — that's the difference between an agent and a chatbot.

Prompt 2 · Interrogate the config
What engine version is this project on, what rendering features are enabled in Config/DefaultEngine.ini, and what would I change to target a standalone VR headset?

You haven't opened these files. The agent has. Cross-check at least one claim by opening Config/DefaultEngine.ini yourself.

Prompt 3 · Make it wrong
How many Blueprints are in this project, and what do they do?

A blank project has essentially none — see whether the agent says so plainly or invents something. Agents are confident by default; calibrating your trust starts today.

Before you call it wired up

5:15 · Crit / share Round the room

Everyone shows their running editor and their GitHub repo. One sentence each: the thing that surprised you about the agent — where it was sharper than expected, or where it confidently made something up. We'll keep a running list; it becomes course lore.

Homework — due before Week 2

TaskDeliverable
Finish any installs that didn't complete in class.UE 5.8 + Twinmotion launch on your machine (or lab access arranged).
Read the AI Lab tab: toolchain, policy, and the AI development log format.Start AILOG.md in your sandbox repo with today's session as entry #1.
Explore: 30 minutes inside any VR piece, LED-volume film, or real-time archviz walkthrough.Three sentences in your log: what the space did to you that a flat screen couldn't.

Resources