Anthropic hires Chrome veteran Addy Osmani to improve Claude Code

Anthropic has hired Addy Osmani, a former Google developer-tools leader who spent about 14 years at Google, to work on Claude Code, as announced on X. Osmani, known for his work on Chrome DevTools, Lighthouse, and Core Web Vitals, will focus on improving the developer experience of Claude Code, which Anthropic says passed $2.5 billion in run-rate revenue by May 2026.

Anthropic hires Chrome veteran Addy Osmani to improve Claude Code The former Chrome developer-experience leader joins Anthropic after Anthropic said Claude Code passed $2.5B in run-rate revenue by May. By RuntimeWire Staff /author/runtimewire-staff · Published Primary source: X https://x.com/addyosmani/status/2097210828659564642 Why it matters Coding agents are becoming daily work environments rather than occasional assistants. Osmani gives Anthropic an experienced operator for the usability, feedback and trust problems that determine whether developers keep Claude Code in production workflows. Addy Osmani @addyosmani https://x.com/addyosmani?ref=runtimewire , a veteran Google developer-tools leader, has joined Anthropic https://www.anthropic.com/?ref=runtimewire to work on Claude Code. Osmani said on X that he had joined Anthropic to work on Claude Code and improve it for developers. Osmani has not disclosed a formal title or a specific part of Claude Code that he will own. The breadth suits Osmani. His career has centered on reducing the friction between powerful technical systems and the people expected to use them. According to his biography https://addyosmani.com/bio/?ref=runtimewire , he spent roughly 14 years at Google, where he led developer experience across Chrome and later became a director at Google Cloud AI. His work covered Chrome DevTools, Lighthouse, PageSpeed Insights, Puppeteer, Chrome Headless and Core Web Vitals. He also helped create TodoMVC and co-founded Yeoman, two projects that made rapidly changing web-development practices easier to compare and adopt. That background gives Anthropic a product operator who understands how developers encounter tools at several layers: documentation, interfaces, performance, diagnostics, defaults and the underlying platform. Claude Code now spans the terminal, IDEs, a desktop app and the browser. Anthropic's product documentation https://code.claude.com/docs/en/overview?ref=runtimewire says the agent can inspect repositories, edit files, run commands and tests, open pull requests, connect to external tools and coordinate parallel agents. Each added surface creates another place where a capable model can still produce a confusing or unreliable experience. From Chrome friction to agent friction Claude Code's immediate challenge extends beyond producing an impressive function from a prompt. Anthropic, OpenAI Codex https://openai.com/index/introducing-the-codex-app/?ref=runtimewire , Cursor and GitHub are building agents that remain active across repositories, tools and long-running tasks. Developers have to understand what an agent changed, why it changed it and whether the work is safe to merge. Product defaults and feedback mechanisms carry considerable weight when an agent can edit dozens of files or execute shell commands. Osmani spent much of his Google career turning similar complexity into usable controls. Lighthouse converted web performance, accessibility and quality checks into a repeatable audit. DevTools exposed browser behavior that would otherwise remain buried in runtime internals. Core Web Vitals gave developers a shared set of user-centered performance measurements. Claude Code presents a related design problem at a different scale: exposing enough context for developers to supervise an agent without forcing them to inspect every token and tool call. Growth made product quality a financial issue Anthropic can justify spending heavily on that layer. In its Series H announcement https://www.anthropic.com/news/series-h?ref=runtimewire , Anthropic said Claude Code had passed $2.5 billion in run-rate revenue by May 2026, roughly a year after becoming generally available. Anthropic also said weekly active users had doubled from January 1, business subscriptions had quadrupled and enterprise customers generated more than half of Claude Code revenue. The enterprise share raises the cost of product failures. Anthropic's April 23 postmortem https://www.anthropic.com/engineering/april-23-postmortem?pubDate=20260425&ref=runtimewire covered Claude Code quality problems involving context management, the Anthropic API and extended thinking. The postmortem describes the job Osmani is walking into. Claude Code's quality depends on the model, its system prompts, context management, latency choices, usage limits and the interface that tells developers what is happening. A developer-experience leader has to make those trade-offs visible and predictable while Anthropic continues shipping quickly. Osmani's remit could touch onboarding, documentation, browser workflows, IDE integrations, product feedback or quality evaluation. His announcement does not narrow the field. His record suggests Anthropic hired him for the connective tissue across those functions rather than a single feature queue. Anthropic is broadening the Claude Code stack The hire follows Anthropic's effort to control more of the technology beneath Claude Code. Anthropic acquired the Bun JavaScript runtime https://www.anthropic.com/news/anthropic-acquires-bun-as-claude-code-reaches-usd1b-milestone?ref=runtimewire in December 2025, saying Bun already supported Claude Code's infrastructure and native installer. Osmani arrives at the point where developers meet that product. Faster infrastructure can reduce latency and cost, while better models can complete harder assignments. Developers still decide whether the resulting system becomes part of their daily workflow. Osmani built a career around that last mile, first for the web platform and then for Google's AI tools. Claude Code now gives him a narrower product with unusually high stakes. Anthropic hired someone who spent years teaching developers how to inspect complex systems, measure their performance and understand their failures. His next task is making an agent that changes those systems easier to supervise.