A quiet corner for technical notes, poems, and fragments of life
Recent entries
Series Runtime Architecture of LLM Agents 1/2 Architecture
From Tool Calls to an Event-Driven Runtime: The Runtime Architecture of LLM Agents
LLM agents should not be understood as a chain of tool calls; as tasks grow longer and environments become more complex, the runtime needs an event-driven architecture for state, concurrency, observation, and recovery.
Series Runtime Architecture of LLM Agents 2/2 Control Flow
Who Calls Whom: Understanding LLM Agents Through CSP and Coroutines
Instead of understanding LLM agents as models calling tools, we can view them as systems made up of communicating processes and coroutines; this perspective better explains event sources, control transfer, suspension, and recovery.
The Culture of the Codebase and the New Paradigm of Coding Agents
Coding agents do not automatically improve a codebase; they enter, learn, reproduce, and amplify its existing culture.