Marcus Chang·Aug 14How Should We Evaluate AI Agents?A practical framework for moving from “the answer looks right” to repeatable evidence about whether an agent actually worksA response icon2A response icon2
Marcus Chang·Aug 10From Prompt to Graph Engineering: More Than Just New TermsEach engineering layer emerged from a new failure mode as AI systems evolved from generating text to taking action.
InTowards AIbyMarcus Chang·Aug 5Loop Engineering for AI Agents: Implementation Beyond TheoryA practical architecture for triggering work, executing tasks, verifying results, eenforcing limits, and improving from failures.A response icon1A response icon1
Marcus Chang·Aug 4Tool Calling Is Not Tool ExecutionThe runtime between a model and the real world is where agent reliability begins
Marcus Chang·Aug 3An AI agent is not the model. It is the system around the model.Why tools, state, permissions, and control loops matter more than another prompt
Marcus Chang·Aug 3The Smallest Useful AI Agent LoopHow messages, tool results, and stop reasons turn a model call into a continuing system
Marcus Chang·Jul 23What Learning AI Agent Architecture From Scratch: A Complete 22-Chapter GuideMany developers are already using Claude Code, Cursor, Codex, and other AI agents.