Google’s developer tools head explains how AI is changing coding

Google

Ryan Salva, Google’s project manager for developer tools, has offered detailed insights into how artificial intelligence is transforming software development in an interview to TechCrunch. Salva, who previously worked at Microsoft and GitHub, now oversees products like Gemini CLI and Gemini Code Assist. In a recent discussion, he shared how his own workflow has been reshaped by AI and what this means for the broader developer community.

Salva explained to TechCrunch that a clear shift occurred around April 2024 with the arrival of new reasoning models like Gemini 2.5 and Anthropic’s Claude 3. These systems introduced strong tool-calling capabilities, meaning they could compile, test, and self-correct code in ways earlier models could not. This changed developer behaviour in measurable ways, with many now treating AI tools as partners rather than assistants.

In his own practice, Salva uses Gemini CLI for side projects but also experiments with competitors such as Claude Code and Codex. He often relies on IDEs like Zed, VS Code, and Cursor not for writing but for reviewing AI-generated output. For professional work, he drafts technical specifications and requirements documents in natural language and then uses AI tools to turn them into functioning code.

He estimates that 70–80% of his development process now happens through natural language interaction in the terminal. Instead of typing out every line, he spends more time articulating requirements and reviewing code for accuracy and architecture. He believes this reflects a wider shift in the industry, where the developer’s role evolves into that of an architect rather than a coder.

Salva argued that this does not make human developers obsolete. Instead, it elevates their responsibilities to higher levels of abstraction. Developers will still need to think critically, break down complex problems, and ensure that solutions are correct and ethical. But manual coding may become less central than it has been for decades.

Looking ahead, he believes that the dominance of IDEs could decline as AI agents handle more of the writing process. Developers may spend more time managing requirements, testing frameworks, and overall system design. This evolution may reshape what it means to be a software engineer, but Salva is confident it will remain a human-led discipline.