Posts
4 postsModel Ergonomics: How Harnesses Shape Agent Capability
The same model can behave very differently under a different harness. This article examines model ergonomics: the fit between models, tools, context, feedback, and execution environments, using Grok on the Codex harness to explore the boundary between generalized and specialized harnesses.
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.
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.