-
Serrino Liu
Tencent
AI Product Expert
Tencent Miora AI Product Expert – Master's degree in Artificial Intelligence from Harvard University, specializing in AI Agent product design, interactive experience innovation, and industry applications in film and content creation.
With hands-on experience ranging from foundational AI model development and data practices—where he collaborated with engineering teams to refine models through product definition—to his current focus on generative AI and Agent-oriented products, he has independently led the design and experience refinement of Agent products. He brings both deep technical understanding and a cross-domain product perspective.
Currently at Tencent's Miora team, he leads product innovation in film/3D and interactive experiences, driving the real-world deployment of AI Agents in content creation and interactive scenarios. He continuously explores the intersection of model capabilities and genuine user needs, committed to transforming cutting-edge AI technologies into intuitive, accessible, and iterable product experiences, and shaping the next generation of human-AI co-creation product paradigms.
Beyond Prompts: How Agents Learn to Create Together
When a user says, "Help me make a short film," it sounds like a simple instruction, but behind it lies the collaboration of multiple Agents – for story, characters, storyboards, images, video, audio, and more. Each Agent may have strong professional capabilities, but stacking capabilities does not automatically produce collective intelligence: they may lack shared context, forget versions the user has already approved, misunderstand each other's work, and not know what they can do autonomously versus what must be returned to human judgment.
Prompts work well for a one‑off generation, but struggle to support the ongoing evolution of a creative work. As AI evolves from a tool that responds to commands to an Agent that understands context, invokes capabilities, and takes action, the design challenge shifts: it’s no longer just "how to get a machine to execute instructions correctly," but "how to help a human and a team of Agents build a shared understanding and develop a collaborative rhythm over time."
This session will share design practices from Tencent's Miora orchestration console, exploring how to transform human‑Agent collaboration into system‑supported capabilities through shared memory, context allocation, action boundaries, and failure recovery. It covers four parts:
1. From Prompt to Project – Why a single instruction cannot complete a creative work
2. Why individual Agent intelligence does not automatically become collective wisdom
3. The orchestration console – Enabling Agents to co‑create
4. From instruction execution to human‑Agent synergy
1、A systematic design framework
2、First‑hand practice from Tencent Miora
3、A new understanding of "warm intelligence"
