Practical ways to plan, inspect and test software made with coding agents, from a small change to a team's development process.
Why read Addy Osmani?
Addy is useful when an agent can generate code but you still need a dependable way to finish the job. His workflow starts with a written specification, breaks changes into small tasks and uses tests, browser checks and frequent commits to keep progress inspectable.
His later lifecycle article explains the surrounding machinery: instructions, tools, context and evaluations all affect an agent's result. Read the workflow for changes you can make today, then the lifecycle piece to understand what a team needs around the model. His current work at Anthropic is on Claude Code.
Start with the original
Selected work
Workflow ·
My LLM coding workflow going into 2026
How do you keep an AI coding session moving toward a checked result? Follow his sequence of specifications, small changes, relevant context, tests and commits.
Engineering essay ·
The New Software Lifecycle
What changes when agents work across the development process? His discussion of the agent's tools, context and evaluations helps explain why choosing a model is only one part of the job.
Projects & roles
agent-skills
His collection of engineering instructions for AI coding agents, linked from his author site.
Beyond Vibe Coding
His book and companion material on AI-assisted development.