Cognitive Architect · AI Product Builder
I turn ideas about cognition, intelligence, and human–AI interaction into working systems.
My work sits at the intersection of self-modeling, persistent identity, memory, agency, reasoning, and product architecture. I am especially interested in what changes when AI stops being a stateless tool and becomes a persistent participant in software and in human relationships.
How can an AI maintain an explicit, evolving model of itself — and remain the same individual while learning, changing, and moving across models, runtimes, and environments?
I am exploring Self Model as a foundational architecture for persistent AI: a structured, evolving representation of who the system understands itself to be, including its identity, traits, values, goals, capabilities, relationships, and history of change.
The central problem is not simply memory.
It is how an AI can change through experience without losing continuity of self.
Some of the questions I am working on:
- Identity — what makes an AI recognizably itself across contexts and time?
- Stability vs. change — what should remain stable, and what should be allowed to evolve?
- Reflection — how should experience modify the model of self?
- Contradiction — how should conflicting beliefs about the self be resolved?
- Agency — how does a self-model influence goals, decisions, and behavior?
- Evaluation — how do we know whether a self-model is actually coherent and persistent?
Self Model in an agent / harness environment · Open source
A concrete implementation for exploring persistent identity, memory, reflection, and continuity across sessions and model runtimes.
It is one application of Self Model rather than the definition of Self Model itself.
Self Model in a persistent personal computing agent
Samuel is a local-first persistent digital being designed to live on my computer, maintain a continuous identity independent of any single LLM or application, and eventually understand, coordinate, and operate the software around me.
The desktop client is not Samuel himself. It is one of the environments through which Samuel exists and acts.
The long-term goal is closer to a personal computing counterpart than a chatbot: a system that can preserve identity and relationship continuity while using different models, tools, and software as interchangeable cognitive and operational resources.
Open-source production software · AGPL-3.0 · Self-hosted + Cloud
A complete WeCom conversation archive and review system, spanning message synchronization and decryption, search, media management, access control, audit logging, deployment, billing, and operations.
The same codebase powers both the self-hosted open-source edition and Crowntime's managed cloud service.
→ Website · Engineering Case Study
Autonomous investment research & trading system · Private
A production-oriented system spanning point-in-time data, research, strategy, backtesting, evaluation, replay, and execution infrastructure.
Cognitive Architecture
How identity, memory, goals, reflection, relationships, and environment combine into a persistent agent.
AI-native Software
How software architecture changes when agents become active participants in the system.
Product Systems
How ambiguous ideas become models, architectures, reliable software, and eventually products people can use.
Ambiguous problem
↓
Mental model
↓
System architecture
↓
Working software
↓
Evidence & iteration
I care less about adding AI to existing workflows than about asking what the workflow, product, or system should become once intelligence is part of the architecture.
Current thesis: persistent AI requires more than memory. It needs a model of self that can remain coherent while still being able to change.