01RESEARCH DIRECTION
PERSONAL AGENTS
Intelligence with continuity.
An agent should carry a coherent identity, understand its person and act across contexts. Memory, learned skills and human relationships must remain distinct—and work together.
A VENTURE BY WYON LIURESEARCH × PRODUCT × COMPANY BUILDING
AI that learns
from experience.
Personal agents. World models.
Recursive self-improvement.
LiveLiva brings these research directions into a human-centered platform: AI that grows with people, connects them and participates in the worlds they create.
THE RESEARCH THESISFIG. 01
01α
Understand the person
PERSONAL AGENTS / HUMAN MODELS
intent informs action
02ω
Model the world
WORLD MODELS / SPATIAL INTELLIGENCE
action produces experience
03Δ
Learn how to learn
CONTINUAL LEARNING / RSI
↳
Evaluated improvements return to the agent
and the world-building system.
Research agenda. Learning gains remain to be demonstrated.
01 / RESEARCH & ENGINEERINGPersonal agents · world models · RSI
Three questions.
One coherent direction.
How can AI understand an individual, act in a persistent world and become more capable through experience? LiveLiva connects these questions in one product and systems agenda.
02RESEARCH DIRECTION
WORLD MODELS
From perception to consequence.
Spatial intelligence connects perception and generation. A world model predicts what an action may change; a persistent runtime determines what actually happened.
03RESEARCH DIRECTION
RECURSIVE SELF-IMPROVEMENT
Improve the way AI improves.
Continual learning must retain and transfer capabilities. RSI goes further: improve the process that produces improvements, and test descendants independently at a matched budget.
02 / THE LIVELIVA PLATFORMFrom research to human value
Personal intelligence.
Shared human worlds.
The product ambition is a network of living worlds: persistent personal AI, real human relationships and places created by people. The first compelling experiences establish the standard for a broader creator platform.
PERSONAL AI
A life that grows with you
Identity, understanding, memory and useful abilities persist across experiences.
HUMAN CONNECTION
Relationships through participation
AI helps people discover shared interests, meet and create a history together.
CREATOR WORLDS
Culture becomes something to do
Local stories, crafts and imagination become playable, social worlds with creator livelihoods.
ONE POSSIBLE JOURNEY
A local story→A shared adventure→A companion & a new ability→Another creator’s world
Illustrative product path. World selection and launch scope are not finalized.The platform hypothesis: continuity makes new worlds more valuable; creators expand what people can experience. Proposed revenue includes paid experiences, personal services and expressive items, with creator revenue sharing.
Explore the product blueprint ↗03 / SYSTEMS ARCHITECTURETarget architecture · research boundaries explicit
A system for experience.
A discipline for improvement.
Nuwa reasons and acts. Fuxi maintains shared state. Pangu creates worlds. World models support planning; learning research proposes upgrades that must earn their way into the system.
- Pangu · createSources → editable, tested world packages
- Nuwa ↔ Fuxi · actAction proposals → verified outcomes → personal experience
- World models · predictState + action → possible consequences for planning
- CL → RSIEligible experience → candidates → independent evaluation → release / rollback
CURRENT EVIDENCE
Local protocol checks: identity continuity, action conflicts, permission revocation and restart recovery. Synthetic accounts; no model calls in these checks.
NEXT PROOF
Evidence scope ↗A compelling first experience and a second independent creator’s world. Validate return visits, cross-world value, production economics and learning gains separately.
04 / FOUNDER & TRACK RECORDFull CV & project evidence ↗
TECHNICAL FOUNDER / AI SYSTEMS BUILDER
Wyon Liu刘怀洋
From the algorithm
to the company.
BSc and MSc in Computer Science, Peking University. Seengene founder & CEO for eight years; later led innovation at Beike and served as QianDing CTO. Experience spans game engines, spatial computing, model engineering, AI characters and company building.
Spatial AI & developer platforms
Spatial reconstruction, localization, generative editing and mixed rendering. Arto reached 50,000+ developers.
Background & evidence ↗Interactive worlds & agents
Dozens of autonomous AI characters with memory and behavior, real-time interaction and AR/VR connectivity.
Background & evidence ↗Company building
Eight years as founder & CEO: nearly RMB100M raised, a 100-person team and RMB tens of millions in revenue.
Background & evidence ↗Selected past work, as documented in the public CV. These outcomes belong to prior projects and companies; they are not LiveLiva traction.
A full-stack perspective on intelligence.
L0 → L5L0
Foundations
Computer vision × graphics. Algorithms, rendering engines and technical business judgment.
↗L1Platforms
Large-scale connections, spatial infrastructure and end–cloud systems.
↗L2Models
Reconstruction, generation, spatial representations and model adaptation.
↗L3Evolution
Agent engines, skill orchestration and evaluated improvement methods.
↗L4Interaction
Real-time AI characters, spatial interaction and feedback-driven learning.
↗L5AI for life
Personal agents, A2A relationships and worlds people create together.
↗05 / SELECTED MATERIALSA short introduction. A deeper conversation.
Explore the work
behind the thesis.
Start with the brief. Review the architecture and founder background. The longer-form ideas and concept film show the product ambition in context.
01BRIEF↗02BLUEPRINT↗03ESSAY / AUDIO↗04BACKGROUND↗
Founder brief
The product direction, founder and next-step opportunities.
Product & systems architecture
How personal agents, world models and creation connect.
CaptainCast · latest conversation
New: 64 minutes on AI companions, worlds and RSI. Both episodes available.
Career & selected work
Engineering, products and company-building experience.
LIVELIVA / CONCEPT FILMNot gameplay
A visual introduction to the intended experience. Product and research status are described in the blueprint.
06 / BUILD THE NEXT CHAPTERAI BUSINESS LEADERSHIP · INTERNAL INCUBATION · INVESTMENT
An ambitious direction.
A substantive conversation.
I’m open to joining a company to lead a new AI product or business, with responsibility for product, engineering and building the team. LiveLiva is the early product direction I’m developing, and a basis for discussing internal incubation.
For investors, I’m also exploring early-stage funding to build LiveLiva independently. These are distinct paths; the next conversation should establish the role or investment mandate, the team and the resources available.
wyonliu@gmail.com ↗LinkedIn ↗