Addy Osmani is a senior engineering leader at Google Cloud AI, focused on AI-assisted engineering, developer tools, and scalable web performance. Author, speaker, and longtime Google engineer with 25+ years of experience shaping modern developer experiences.
25 articles from this blog
Explores the concept of software factories using AI loops, comparing human-in-the-loop (light) vs. autonomous (dark) approaches, and the engineering challenges involved.
Analysis of agentic engineering, focusing on engineers owning the outer loop for accountability in AI systems.
Explores how AI agents are reshaping junior developer career paths by automating learning reps, impacting taste and judgment development.
Career advice for engineers in the AI era, focusing on scarce resources like reputation and problem selection over solution skills.
Explores agentic autonomy levels in AI engineering, proposing a two-axis framework of agency and orchestration for multi-agent systems.
Explores how AI is transforming the software development lifecycle, focusing on agent harnesses and context engineering.
Explores how AI coding agents shift engineering focus from writing code to code review, making review the most leveraged skill in software.
Explores loop engineering, a new approach where developers design systems that prompt AI agents instead of manually prompting them.
Explores intent debt as a hidden cost in software engineering, distinct from technical and cognitive debt, and why AI agents can't resolve it.
Explores the 'orchestration tax' in AI agent workflows, comparing human cognitive limits to concurrent system bottlenecks.
Explores how relying on AI for coding without active learning degrades skills, backed by research showing cognitive debt.
Explores 'cognitive surrender' where AI output replaces independent thought, with specific risks for software engineers.
Agent Skills project enforces senior engineering workflows like specs and tests for AI coding agents, preventing shortcuts to 'done'.
Explores 'comprehension debt,' the hidden cognitive cost of over-relying on AI-generated code, which erodes team understanding and long-term maintainability.
A Google engineer shares 14 lessons on teamwork, decision-making, and focusing on the right problems, based on 14 years of experience.
Claude Code introduces agent teams (swarms), enabling parallel AI agents with specialized roles to coordinate on complex coding tasks.
Distinguishes reckless 'vibe coding' from disciplined, AI-assisted professional software engineering, proposing 'agentic engineering' as a better term.
A guide to setting up autonomous AI coding agents that work in continuous loops to write, test, and commit code automatically.
A guide to writing effective specifications for AI coding agents, covering principles like starting with a high-level vision and breaking down tasks.
Explores how managing multiple AI coding agents parallels tech leadership, shifting focus from code generation to orchestration and review.
