SHANGHAI, September 23 — Xisiid Intelligence, an AI neo-lab chasing what it calls next-generation artificial intelligence paradigms, has closed a RMB 25 million (about US$3.5 million) seed round, the company said, timing the announcement to its first public appearance at the Shanghai Pujiang Innovation Forum. The startup is led by founder and CEO Dr. Chua Yam Song. The new capital will primarily support accelerated research and development of its proprietary Brain-Inspired Self-Evolving Foundation System, advance its core memory-centric architecture, and expand co-creation initiatives within knowledge-intensive professional fields.
Here is the shift Xisiid is betting on. The company positions its technology as a departure from standard conversational large language models that generate content on demand.
Instead, it is pursuing what it calls “Own Intelligence” — systems designed to reliably complete complex, long-horizon tasks and to build reusable organizational memory over time. That is a different target than the one most of the industry has been aiming at. Xisiid’s stated focus on long-horizon tasks and organizational memory places it in a segment of AI development that seeks to move beyond single-turn text generation toward systems that retain and reuse experience across repeated work.
Two layers, one architecture
The foundation system runs on a dual-layer cognitive architecture. System 1 is described as a high-speed execution layer that uses a proprietary latent exchange protocol for direct vector-based interaction between modules, avoiding the delays of text decoding. System 2 is a task-centric hierarchical memory layer that captures reusable workflows, decision heuristics, and experience from previous executions.
Deployment matters here too. The company said the system supports on-premise deployment so that mission-critical workflows and organizational knowledge remain completely within enterprise environments, protecting enterprise data assets.
Xisiid frames its core framework as Brain-inspired Intelligence for Artificial Intelligence, with a primary paradigm described as memory-centric and a cycle of Execution, Feedback, Generalization, and Consolidation.
What to watch
The RMB 25 million round is a seed-stage investment, a category typically used to fund early product and research development before a company pursues larger venture rounds. Xisiid did not disclose the identities of the participating investors in the information available. Whether the approach delivers on those claims will depend on research progress and enterprise adoption, both of which the company says the seed funding is intended to accelerate.
Watch this space.
