Facilitator detail for Day 1. Audience: intermediate+ devs who already know Three. Everything runs WebGPU on fiber v10. Code shown here is illustrative — check APIs against current docs when building each chain.

The frame for the day: by 5pm everyone has rebuilt the Paris site and touched the full scope of R3F. The site is the spine — name it that at the top of the day and keep saying it. Every block ends with "add it to the spine." There's no dedicated spine lesson; its mechanics (multiple canvases, rendering only what's in view, onFramed) come up naturally in L2, the state block, and the finale.

How blocks run: 15–20 minutes of "watch me build it" (30 max), then you're on the floor helping while people work through the written steps. Anyone behind can git switch to the current step and rejoin. Agents are welcome — people should lean on them for gaps, as long as they're the ones typing and fixing.

PM block shape: ① the concept on its own → ② quick mini-demos of the sub-ideas → ③ a capstone that puts it in the site.

Who's in the room, and how we respond

What we expect people to have: solid JavaScript, the Three vocabulary (scene, camera, mesh, materials, lighting — they've built something with it), git basics, and passing familiarity with React — they've seen components and hooks even if they don't write React daily.

What we don't expect: any R3F experience, WebGPU or TSL, shader authoring, physics engines. That's what the day is for.

When someone's missing a piece, the playbook is the same: help them at their seat, get them to a working checkpoint, and keep the room moving. The day never stops for one gap.

• Weak React — they'll feel it at components/props in L1 and again in the state block. Help them through it, point them at the React starter from the pre-work, pair them with a neighbor if it persists. We support them all day, but we don't re-teach React to the room.
• Shaders (the TSL block) — expect people to fail here, and say so from the stage. That's normal; shaders are hard. Get them to the working checkpoint, then handwave forward: "this is exactly where your agent earns its keep, and the written guide goes deeper when you have an evening." The capstone checkpoint means nobody's stuck staring at a black screen.
• State — React folks will coast on part one and hit something new at transient subscriptions. That's the fun part; let them wrestle a little before rescuing.
• Physics — fast show by design. Nobody needs mastery; anyone who wants depth gets the stretch set and their agent.
• Machine trouble — hosted capstone on GitHub Pages, follow along there, fix local at the overflow/break.

Morning Lesson 1 — Build the Hero

Hook: the finished hero is on screen when people walk in — the Eiffel Tower under a slow orbit. The lesson is rebuilding it, explaining the choices as you go.

Slot: 09:35–10:30, right after the setup gate.

Covers: the R3F mental model · components and props · useGLTF and Suspense · cameras, controls, backgrounds.

By the end people can:

• Take three.js they already know and write it as JSX — the tree is the scene graph, nothing gets new-ed or add-ed by hand.