- From: 陳信屹 <tyson@slashlife.ai>
- Date: Mon, 10 Nov 2025 22:15:00 +0800
- To: public-cogai@w3.org
Thank you for the welcome. I would like to briefly outline our current research direction and its possible relevance to the Cognitive AI CG. I am investigating the design of an agent-native operating environment in which semantic structures are treated as part of the execution substrate rather than solely as representational artefacts. In this model, ontology operates as an execution constraint at the OS layer, governing agent identity, delegation boundaries, and admissible behaviours. Agents execute as isolated processes with DID/VC identities, semantic contracts, and an append-only episodic ledger. We are not yet implementing full reasoning frameworks; LLMs are used only as partial inference modules. Our current work is focused on the system prerequisites for cognitive behaviour, including: — architectures for episodic, working, and narrative memory — mechanisms for metacognitive signalling and attention regulation — responsibility and traceability models for multi-agent execution — methods for mitigating LLM context interference (confusion across tasks, narrative drift) through OS-level context isolation and memory management These form the lower-level structures upon which higher-order reasoning formalisms (e.g., chunks, rules, or PKN) could be integrated. The broader objective is to explore a next-OS computation model suitable for multi-agent coordination, digital twins, and embodied/immersive contexts. For reference, a working draft of the agent-communication ontology used in this environment is available at: https://github.com/slashlifeai/agent-ontology/tree/main I would be glad to contribute to discussions on semantic execution, identity/delegation models, episodic memory, and LoA alignment between cognitive reasoning and system-level semantics.
Received on Monday, 10 November 2025 14:37:28 UTC