Addy Osmani's Workflow

2026-05-14 · page · source summary

Addy Osmani's comprehensive guide to integrating AI coding assistants into professional development. The core argument: classical software engineering practices become more critical, not obsolete, when you add AI. Start with a spec.md (requirements, architecture, data models, testing strategy), work in iterations (plan + implement + test + commit), and treat AI output like junior developer code -- always review, always test.


Key Quotes

"AI coding assistants are incredible force multipliers, but the human engineer remains the director of the show."

"The LLM is an assistant, not an autonomously reliable coder...I am the senior dev."

"At Anthropic, engineers adopted Claude Code so heavily that ~90% of the code for Claude Code is written by Claude Code itself."

"Think of an LLM pair programmer as over-confident and prone to mistakes."

Key Themes

#agentic-coding #spec-driven #testing #code-review #iteration #context-packing

The spec.md approach is the most transferable idea here. It's "waterfall in 15 minutes" -- use AI to rapidly iterate on requirements before touching code. The emphasis on examples as context (not just instructions) is practical and underappreciated. His "model musical chairs" suggestion -- switching between Claude, Gemini, and others when one stalls -- is pragmatic but adds cognitive overhead.

The iteration pattern (plan + implement + test + commit as one unit) connects to Planning With Files's approach to maintaining state across sessions. Both are trying to impose structure on inherently chaotic AI-assisted development.

Critical Analysis

This is the responsible, professional-grade take on AI coding. Almost everything here is sound advice. The tension is in the closing: "No matter how much AI I use, I remain the accountable engineer." That's true today, but Five Levels from Spicy Autocomplete to the Dark Software Factory suggests we're heading somewhere that makes it less true. Osmani's workflow is solidly Level 2-3 in Shapiro's framework. The question is whether the discipline he advocates scales to Level 4+, or whether it becomes impossible to maintain human oversight at higher automation levels. See also The Next Two Years of Software Engineering (his companion piece on the macro trends) and 14 More lessons from 14 years at Google for his organizational thinking.


Sources: Addy Osmani's Workflow Last updated: 2026-05-14