From “Lobster” to Ontology: Unveiling the Next Wave of Self‑Evolving AI Agents and Data Governance

On June 24‑25, 2026, DataFun hosted the DACon conference in Shanghai, attracting more than 8,000 technical developers, managers and industry experts. The two‑day event featured 50 talks across five parallel tracks, covering deep technical insights and broad applications.

Keynote 1 – Self‑Evolving Agents : Professor Xiao Yanghua of Fudan University presented “Technical Breakthroughs and Evolution Trends of Self‑Evolving Agents,” focusing on the open‑source “Lobster”‑class framework GenericAgent, scheduled for public release on 2026‑01‑11. He highlighted its early launch compared with OpenClaw, fully self‑developed codebase, comprehensive environment interaction and task execution capabilities, a minimalist architecture of just over 3,000 lines, and the “seed philosophy” that lowers the entry barrier for agents.

Keynote 2 – Agent‑Ready Big Data Infrastructure : Wei Bowen, solution lead at Alibaba Cloud Intelligent Group, described how agents are reshaping software development, content creation and customer service. He detailed Alibaba Cloud’s system‑level breakdown of multimodal data, vector retrieval, data timeliness and security governance, positioning the platform as the foundation for Agent‑Ready big‑data AI infrastructure.

Keynote 3 – Ontology in Banking : Yu Haohan, General Manager of Data Management at Shanghai Bank, explained the challenges of knowledge‑base construction, intent‑recognition bottlenecks and hallucination control in large‑model engineering. He introduced a three‑layer ontology value model—semantic, knowledge, and constraint layers—that forms a trustworthy AI knowledge base, reducing hallucination risk and improving intent‑recognition accuracy.

Keynote 4 – From Data Consumers to Agents : Du Junping, founder and CEO of Datastrato and initiator of Apache Gravitino, traced a decade of solving “how humans use data more efficiently.” He argued that the next wave of massive data consumption will be driven by AI agents, requiring a new protocol layer—Agentic Data Protocol (ADP)—to handle discovery, authorization, provenance and compliance.

Keynote 5 – From AR to Agents : Yang Zhou, Global Ecosystem R&D Lead at Rokid, presented the evolution of the AI‑AR ecosystem, showcasing Rokid AR Lite, AR Studio Pro and AI smart glasses. He described a no‑code “Lingzhu” platform that has already launched over 300 agents covering daily life, productivity tools and interactive games, and outlined how AI‑AR integration fuels digital entertainment, industrial simulation, digital twins and remote expert guidance.

The conference schedule also included specialized sessions on Agent‑Ready data architecture, Agentic RAG, AI‑driven data development, LLM‑powered research, OpenClaw commercialization, and cost‑optimization for large‑model applications. Partners such as Alibaba Cloud, NVIDIA, Tencent Cloud, and many others demonstrated their latest products and engaged with attendees in the exhibition hall.

Looking ahead, DACon organizers announced the upcoming Agentic AI Summit in Shenzhen (July 7‑8, 2026), promising concrete, reproducible solutions for AI agent deployment.