Google just gave K-12 schools a powerful lever: Gemini, Gemini Notebook, and Search wil...

When Schools Reject Chatbots, They're Asking for Spatial Computing — Here's Why I Built for That Moment

I stood in a middle-school hallway last October, watching a seventh-grader named Marcus try to explain a glucose molecule to his lab partner using a flat Chromebook screen. He tapped, pinched, scrolled. The 2D diagram refused to rotate. His partner shrugged. "I don't get it."

That afternoon, the district's tech director emailed: "We're blocking ChatGPT. Kids are pasting essays. But we still need something that helps them see the science."

She wasn't asking for another LLM wrapper. She was asking for spatial computing — without using those words.

I'd spent six months wrestling with WebXR, Three.js, and the painful reality that most school hardware tops out at integrated graphics. My prototype ran at 12 fps on a 2018 iPad. The shader compilation pipeline was a mess; I'd over-engineered the physics engine for molecular dynamics nobody had asked for.

Strip it down. Keep the molecule viewer. Cache the GLTF assets locally. Use WebGPU where available, fall back to WebGL 2 with a reduced instruction count. Cut the physics simulation — pre-bake the conformational states. The bundle dropped from 4.2 MB to 680 KB gzipped. First meaningful paint on that iPad: 1.8 seconds.

Marcus came back in January. He grabbed the headset — a $299 Meta Quest 2 the district bought on grant money — and pulled the glucose molecule into his hand. Rotated it. Pinched the glycosidic bond. "Oh," he said. "The oxygen bridges the rings."

His partner leaned in. "Can I try?"

The tech director's next email: "What else you got?"

I'm building the cell-division module next. Mitosis in your palm. Meiosis on the library table. If you're watching schools hit the chatbot wall and wondering what comes after — this is it.

spatialclassroom.io