Noodle

The brief

Sketch to 3D in one
spatial flow.

MIT Reality Hack 2026 open brief: build something meaningful for spatial computing in 36 hours. The team chose to solve a real problem for creative professionals. The constraint was the hardware: Snap Spectacles Gen 5, hand tracking only, no keyboard, no mouse.

Instead of building another AR productivity tool that digitised an existing desktop workflow, we asked what a creative workflow looks like when it starts in physical space. The answer was a node-based spatial canvas where a sketch on your real desk becomes the first input, your voice becomes the prompt, and a 3D model sitting on that same desk is the output.

No app switching, no file management, no keyboard. The entire idea-to-object pipeline in one continuous spatial flow.

Event

MIT Reality Hack 2026

Platform

Snap Spectacles Gen 5

Role

Creative Director · Lead XR Developer

Awards

Founders Lab Track Prize · Best Use of Spatial AI

Year

January 2026

Deliverable

Mixed reality creative workbench with node-based spatial AI pipeline.

Process

Problem before idea.
Interface as logic.

The problem before the idea

Most hackathon projects start with the technology. We started with a friction point: creative professionals switch between an average of 10 applications to take a sketch to a 3D concept. That number became the brief we gave ourselves. Every design decision after that was tested against one question: does this reduce the switches or add one?

The wire system as interface

The node-and-wire interaction was not just visual. It was the logic of the tool made physical. We built a custom wire renderer in TypeScript so that dragging a connection between two nodes actually triggered the pipeline. The act of connecting became the act of creating. That decision unified the UI and the backend into a single gesture.

Designing for latency

Generative AI takes 10 to 30 seconds to return an output. In a keyboard-based tool, that is a loading bar. On Spectacles with no screen to retreat to, it would have broken the experience entirely. We designed visual feedback states that made the wait feel like the system was thinking alongside you, not making you wait.

The result

Two prizes, one flow.
Constraints that paid off.

Winner Founders Lab Track Prize at MIT Reality Hack 2026

Winner Best Use of Spatial AI, sponsored by Snap Spectacles

Winner Snap Spectacles Community Challenge

Friction, not features, produces tighter concepts

The "Toggle Tax" insight gave the whole project a spine. Every feature we considered could be evaluated against one test: does it reduce context switching or not? That constraint eliminated scope creep and kept the concept coherent under hackathon time pressure.

Hardware constraints force better design

No keyboard on Spectacles is a limitation that became a design principle. Voice-first and gesture-first interaction is not just accessible, it is faster for the right tasks. The constraint produced a more considered UI than an unconstrained brief would have.

Latency is a UX problem, not a technical one

The 10 to 30 second generation wait is a fixed technical reality. Treating it as a UX problem to solve rather than a spec to apologise for changed the output entirely. The experience felt fluid not because generation was fast, but because waiting was designed.

Physical and digital can share one logic

Grabbing a real sketch off a real desk and pulling it into a spatial node graph sounds complex. It worked because the interaction metaphor was consistent throughout. When the metaphor holds, the learning curve collapses.

Sketch to 3D in onespatial flow.

Spatial nodes,real desk, AI output.

Problem before idea.Interface as logic.

Two prizes, one flow.Constraints that paid off.

Sketch to 3D in one
spatial flow.

Spatial nodes,
real desk, AI output.

Problem before idea.
Interface as logic.

Two prizes, one flow.
Constraints that paid off.