Engineering systems that stay dependable under real constraints.
Original, bilingual essays by Corn Han (韩朋朋) on production AI agents, real-time full-stack systems, and secure Web3 infrastructure.
From Demo to Production: Engineering Reliable AI Agent Systems
A practical architecture for tool contracts, orchestration, memory, observability, evaluation, and human control in production AI agents.
Designing Responsive Full-Stack Systems for Real-Time FinTech
An engineering model for market-data ingestion, state isolation, rendering budgets, backpressure, recovery, and cross-terminal consistency.
Engineering AI Agents for Web3 Without Surrendering Control
A security-first architecture for transaction context, simulation, permission boundaries, human approval, and reproducible AI-assisted Web3 workflows.
Written from systems work, not trend summaries.
These essays turn first-hand engineering patterns into reusable architecture: explicit boundaries, observable execution, controlled failure, and decisions that remain understandable to users and operators.
