23 Daily — According to 36kr.com, Li Shimiao, a former Asia-Pacific member of Disney, founded the interactive intelligent platform company SYNAPUFF in 2026, focusing on AI Native terminal interactive intelligence. The company recently completed a seed round financing amounting to several million RMB, with investors including FIRST RULE VENTURES and Pudong Ventures. Li stated that SYNAPUFF aims to enable intelligent agents not only to know what they can do but also to determine when, where, and how to interact and dynamically adjust their behavior based on user responses, surpassing traditional user-initiated command models.
Founding Team Background and Technical Foundation
Li Shimiao’s career spans entertainment experience, investment, and human-computer interaction. She previously worked at Disney’s Asia-Pacific headquarters, participating in multiple classic IP derivative projects. Later, she joined Pixieland Technology as a founding team member, involved in systemic research and investment in multiple hard-tech unicorns. In 2021, she founded Ant Touch Technology focusing on tactile feedback models and multimodal interaction technologies serving clients such as LVMH, Kering, and Audi, with investment from Sequoia China. Starting in 2025, she conducted research at the Human-Computer Interaction Lab of Hong Kong University of Science and Technology (Guangzhou) and launched the new SYNAPUFF venture in 2026. Co-founder and Chief Scientist Dr. Mingming Fan is an associate professor and doctoral supervisor at HKUST (Guangzhou), with long-term AI research in human-computer interaction and human-intelligence collaboration. He was listed among the world's top 2% scientists in the 2026 global ranking.
The “P Framework” Breakthrough in Traditional Human-Computer Interaction
Traditional human-computer interaction relies on user-initiated commands and device responses. With AI becoming a native capability of terminals, intelligent agents now have the potential to continuously perceive environments, understand contexts, and proactively provide services. SYNAPUFF focuses on improving the 'emotional intelligence' of agents, enabling them to comprehend interaction contexts, determine interaction boundaries, and dynamically adjust behavior. For example, a robot can detect if the user is in a meeting and avoid interrupting to offer water. The company’s 'P Framework' shifts from a linear 'perceive–understand–act' process to an interactive loop of 'probe–receive feedback–decide–receive feedback–decide.' Agents can probe user intentions through subtle visual or auditory cues and decide subsequent actions based on responses. The framework includes three main components: interactive dynamic modeling, personality strategy selection, and non-verbal behavior generation, integrating existing visual and emotion recognition with native non-textual language gesture generation to adapt across terminal forms. The company is currently training related models.
Data-Driven Interactive Intelligence Training and Industry Focus
Data is central to SYNAPUFF’s model training. The company plans to gather data from controlled experiments, public media datasets, and real-world deployments to compare various agent actions and user feedback, refining interaction policies. Unlike traditional models focusing on task completion, interactive intelligence emphasizes the appropriateness and continuous optimization of behaviors. This technology addresses challenges arising from foundational large model companies extending AI capabilities to smart terminals, focusing on understanding user intent, timing interactions, and responding appropriately—forming a new competitive frontier in AI terminal applications.
Editor's take: New Opportunities in Interactive Intelligence for Entrepreneurs
As AI technologies increasingly integrate with intelligent terminals, SYNAPUFF’s 'P Framework' offers a dynamic, context-aware interaction approach closer to natural human exchanges. This departure from conventional command-response modes points toward more humanized and context-sensitive intelligent devices. Entrepreneurs should monitor such technology’s applicability in consumer robotics, service industries, smart homes, and multimodal interaction sectors.
To validate the technology’s effectiveness, teams can deploy intelligent agents in specific terminal scenarios, using A/B testing to assess improvements in interaction naturalness, contextual adaptability, and user satisfaction pre- and post-implementation of the 'P Framework.' Data collected can further optimize interaction strategies, establishing a closed feedback loop—critical for evaluating commercial viability and user acceptance of interactive intelligence.