Blog
Notes from the crew
Short, honest takes on AI agents and the harnesses that run them.
Claude Code vs Codex vs OpenCode: which agent for which job?
We run all three coding agents daily. Claude Code wins on design and judgment, Codex on code generation and review, OpenCode on cost and model freedom. Here's the honest breakdown of when to use each.
Read itWhat is an AI agent harness?
An AI agent harness runs coding agents like Claude Code and Codex toward a goal you set: it picks the agent, checks the work, and loops until it's done. Here's how that differs from a chatbot, a copilot, and a framework.
Read it