Skip to Main Content

Problem Buying furniture online for a small room is guesswork. A grad student setting up a gaming corner in a shared apartment can't tell from product photos whether a desk will fit, whether the chair will block the door, or whether the whole setup stays under budget. Most mistakes only show up after delivery: a desk that stops the door from opening, or a cart that drifted over budget. What Ethyneon does You photograph your room and film the door opening. Ethyneon turns the photo into an editable 3D room, reads from the video which way the door swings, and fills the room with real store products that fit your budget and style, all by voice or text. Drag a desk into the door's swing and it catches the conflict and offers a checked spot nearby. You pick what to buy, review the cart, and pay on Shopify's hosted checkout in test mode. Nothing is bought by voice; only on-screen buttons create a cart. Why it needs Gemini 4 and real media The photo and video carry what a text description loses: wall sizes, where the door is, which side it's hinged on, and how far it swings. Each model has one job: Neon: builds the room from the photo, reads the door from the video, and plans each request as a checked workflow Hydrogen: understands spoken requests and answers out loud Decision 1: runs six fast checks (photo quality and privacy, door clip, request routing, which item you mean, catalog prompt injection, checkout audit) Argon and Carbon Managed Agents: Argon proposes layouts in one sandbox, Carbon tries to break them in another, and the server re-checks every claim Proof it works Everything was tested live against the real APIs: Door clip: a fast check answers in about 2 seconds, down from 155 seconds; Neon verifies in the background. Real failure we found and fixed: Neon once said a door was hinged on the left at 0.85 confidence when it was hinged on the right. We added an independent Decision 1 check, and the hinge is now only accepted when both agree. Otherwise the app asks you to tap. Prompt injection: the check caught the one product description written to manipulate the AI and passed the clean ones. Test payment: on our Shopify dev store, card 1 was approved (order in Shopify) and card 2 was declined, with the total matching the approved amount to the cent. Built at the event All code was written today: the AI and agents layer, the fit solver, the 3D web app, the Shopify cart and checkout path, the shared team contracts, and 54 automated tests. The test room media was generated with Nano Banana and Gemini Omni and is labeled as such. The product catalog in the main demo is a labeled test catalog; the live payment runs on our Shopify dev store in test mode. Tracks: Element 119 (media is the core input) and the Shopify bonus.

Team

mitran22rKarish2026shrutisridhar4

More from Gemini 4 NYC Hackathon

We design an agentic system to automatically create + manage intelligent AI opponents/NPCs in video games. The suite of agents analyzes user gameplay…

Adaptive NPCs project preview

Manually routing PCBs is time-consuming and cost-inefficient. Current best algorithms (freerouting, PCBRouter) score very low on PCBWorld's hardest PCB…

routr project preview

Heliocam is a camera app where Gemini controls the camera. It watches what you're pointing at, sets the ISO, shutter, focus, white balance and zoom for you…

heliocam project preview

Repairs and assembly fail because manuals are lost, vague, missing, or nonexistent, so people break things or throw them out. KLIK turns a photo of any object…

Klik project preview

Iron Pulse turns your laptop into a 3D boxing ring. Fight an adaptive AI rival in real time using just your webcam. Throw real punches and lean to dodge…

Iron Pulse Arena project preview

Text your raw footage to an iMessage number and get a finished, trend-informed Reel back in the same thread.

Sticky project preview