coco

cocoverse: a little universe for you and your pet. Live at https://www.cocoverse.app Your pet will never tell you who they are. Pet "AI" today is either a chatbot that knows breeds, or a filter on a photo. cocoverse builds your pet's AI twin the only honest way: from what the animal actually does. Every photo you share becomes a collectible card, a Memory Print. Muse Spark reads the moment into short notes and a line in the pet's voice; Muse Image paints a new picture of your pet in that moment, in a blind-box style; SAM 3.1 cuts the pet out so rare cards have real depth. You add the story the photo can't show, and your words outrank anything the model guessed. The twin grows only from those cards. It answers as your pet, in your language, and names the cards each answer rests on. Asked about something it has never been shown, it says so and suggests the photo that would teach it. It never makes a memory up: an answer can only cite cards that exist, a single photo never becomes a habit, and general breed knowledge is labelled as such. Under every answer it shows what it did: how many moments it looked through, what it checked, which cards it cited. Around it, a small world: Memory Points and daily packs; series that follow who your pet is; a portrait that sharpens from a mosaic to a painting as evidence grows; a memory graph of everything it knows, every link resting on real cards; a 3D character of your pet that walks over and reacts. Around us checks today's weather, whether a parked car is safe right now, and the local rules with their sources, so when you ask your pet along to the shop it checks first. Friends: a sitter can talk to the twin through a link, and before two pets meet, their twins meet first (agent to agent) and both owners get one honest note. Faces are blurred on the device before anything is uploaded. No health, food-amount, medicine or training advice. Any pet. Everything in the repository was written during the hackathon; Coco is a real dog, Biscuit and Mochi are demo pets whose photos were made with Muse Image.