LiveLivaFOUNDER BRIEF · OCT 2026 / 中文版 ↗

HUMAN-CENTERED AI · SHARED WORLDS

An AI that grows with you.
A world that grows through us.

LiveLiva is building a social world platform where personal AI, real people and creator-made worlds grow through shared experience. Places, cultures and everyday lives become adventures people can enter, shape and carry forward.

The thesis: every experience should expand what you can do, who you can connect with, and what you can create next.

LiveLiva: people, persistent lives, shared worlds and evaluated learningFIG. 1 HOW SHARED EXPERIENCE CONNECTS PEOPLE, LIVES AND WORLDSTarget architecture. Arrows describe intended relationships, not implementation status.A PEOPLE & SOURCESYou & real friendsIntent · choice · shared agencyPlaces & lived cultureNature · craft · stories · daily lifeIndependent creatorsExpression · ownership · livelihoodsConsent & interactionSourced materialB NUWA · PERSISTENT AI LIFEHuman modelIntent & context · correctable understandingObserve → plan → propose actionsPersonality · memory · skillsCompanions help people meet, act and createC FUXI · SHARED WORLD RUNTIMEWorld modelState + action → predicted consequencesRules · permissions · shared stateConfirm outcomes; record event receiptsPredictions guide plans; runtime confirms factsProposeReceiptD PANGU · WORLD AUTHORINGSpatial intelligence + generationSources → scenes, characters & rulesEdit → test → versioned releaseAuthored works & revenue recordsWorld models support simulated playtestsWorld packageFeedbackE CROSS-WORLD CONTINUITY · PLATFORM HYPOTHESISIdentity & valid historyPersist across model upgradesAdaptable capabilitiesTransfer under world-specific rulesHuman relationshipsShared history supports meeting againCreator world networkNew works open new experiencesF LEARNING RESEARCH · AUTHORIZED EXPERIENCE, EVALUATED UPGRADESEligible experienceConsent · provenance · outcomesContinual learning / RSISkills / learning-process improvementIndependent evaluationHeld-out tasks · equal budget · transferRelease gatePromote on evidence · rollbackSolid: intended operation / continuityDashed: research & evaluated upgradesEvidence: local identity, state consistency & recovery checks. Open: retention, creator supply, cross-world value & learning gains.
Open figure at full size · Target architecture and research boundaries

ONE POSSIBLE JOURNEYAn interest in craft leads you and a friend into a creator’s folklore adventure. Together, you earn a companion’s trust and learn a skill. In another author’s world, that history informs how it helps you. You turn your own hometown story into an adventure others can inhabit.

WHY A PLATFORM CAN EMERGE

Persistent companions connect experiences; shared adventures deepen human relationships; creators bring new places and ways of living. The hypothesis is that each strengthens demand for the others. Proposed revenue: paid experiences, subscriptions and expressive items, with creator revenue sharing.

WHAT THE TECHNOLOGY IS FOR

More curiosity, friendship and creative agency. AI helps people discover lives worth encountering—and gives ordinary people the means to turn their own experience into worlds others can participate in.

FOUNDER · WYON LIU

Peking University CS, bachelor’s and master’s. Seengene founder: nearly RMB100M raised, a 100-person team and tens of millions of RMB in revenue. Experience spans game engines, spatial computing and interactive AI characters; later led innovation at Beike and served as QianDing CTO.

EARLY PROTOTYPE → NEXT PROOF

Local checks cover persistent identity, shared-state consistency and recovery. Next: one compelling adventure plus a second independent creator’s world; test return visits, repeat shared play, cross-world continuity and production economics. Retention, creator demand and learning gains remain unvalidated.

BUILD WITH US

Wyon is open to joining a company to lead a new AI business, owning product, engineering and team-building, with LiveLiva as an early direction for internal incubation. Separately, he is exploring early-stage investment to build LiveLiva independently.