SeeFoodTHE FLY EDITIONSHAZAM FOR FOOD. SORT OF.
BIG VISION. TINY BRAIN. LIVE DEMO

HOTDOG. NOT HOTDOG.

The world’s most specific food app, straight out of Silicon Valley. Now operated by a fly.

FLY IN CONTROL100% autonomous play
▣
LIVETHE GREAT LUNCH BREAK
ROUND 1
TOTAL SCORE0
+ 0HOTDOG
× 0NOT HOTDOG
Picnic islandTabletop expedition
EXPLORER MAP
THE INTERN ↓
Hotdogs, or pizza, burgers, donuts and tomatoes. Zero human input.
Booting SeeFood…Onboarding our smallest employee
NO BITES YETHOTDOG OR NOT HOTDOG?
LET THE FLY COOK.
ROUND PROGRESS 0 / 512
CAMERA
WORLD
ADVERSARY
—NEURONS
—CONNECTIONS
—ACTIVE
Loading policyi
No one is driving this demo.
Every move comes from the fly’s brain.
↑ Forward—↶ Turn left—↷ Turn right—
L—
R—
One app. Two possibilities.

It’s a hotdog. Or it’s not a hotdog. That’s the whole business.

HOTDOG +1 NOT HOTDOG −1
Muted
POWERED BY A REAL FLY CONNECTOME · LOCAL BRAIN, LIVE DECISIONSAN UNNECESSARILY AMBITIOUS EXPERIMENT ↗
THE SEEFOOD PITCH

Hotdog. Not hotdog. That’s it.

The pitch comes from HBO's Silicon Valley: an app sold to investors as “Shazam for food” that, when finally demoed, could tell you one thing. Hotdog, or not hotdog. HBO shipped a real version of that classifier the night the episode aired. This is the next step nobody asked for: the classifier is now a fly.

The fly plays this game by itself. Two antenna-like sensors detect food signatures. A recurrent network built on a real fruit-fly connectome turns those signals into forward, left, or right actions.

Collecting a hotdog earns a reward. Other food costs points. Each round starts a fresh arena and resets the brain's memory; the score continues across rounds.

This is a connectome-constrained trained model, not a biological fly or a claim of a complete brain simulation. The panel shows the actual loaded neuron and connection counts. Weights stay fixed during play.

Space Pause / resumeDrag Orbit the tableScroll Zoom in / out

Camera, sound, and pace are spectator controls. There is no human steering.