Google’s Developer Tools Lead Discusses AI’s Growing Role in Coding
Key Points
- Ryan Salva leads Google’s developer tools, focusing on AI products like Gemini CLI and Gemini Code Assist.
- Google’s latest developer survey shows AI tool adoption surged around April 2024.
- Tool‑calling capabilities enable AI models to compile code, run tests, and self‑correct.
- Salva’s workflow relies heavily on natural‑language commands via Gemini CLI, with 70%‑80% of tasks handled by AI.
- Multiple IDEs are used mainly for reading code, not for writing it.
- Future developers are expected to act more as architects, guiding AI rather than hand‑coding.
- The role of traditional IDEs is projected to shrink as AI takes on routine coding tasks.
Ryan Salva, the manager for Google’s developer tools, explains how AI-powered solutions like Gemini CLI and Gemini Code Assist are reshaping software development. He highlights findings from Google’s latest developer survey, noting a surge in AI tool adoption around the release of new reasoning models. Salva describes his own workflow, which blends terminal‑based AI commands with traditional IDEs, and emphasizes the shift toward using AI for requirements drafting and code generation. While he sees the IDE’s role diminishing over time, he envisions developers evolving into architects who guide AI rather than write every line of code themselves.
AI Integration in Google’s Developer Tools
Ryan Salva, who oversees Google’s developer tools portfolio, provides a behind‑the‑scenes look at how artificial‑intelligence tools are changing the way developers work. Central to his team’s effort are products such as Gemini CLI and Gemini Code Assist, which enable developers to issue natural‑language commands from the terminal to generate specifications, write code, and manage revisions.
Survey Findings Reveal Rapid Adoption
Google’s annual developer survey this year focused heavily on AI usage. The data showed a clear spike in adoption that coincided with the launch of new reasoning models, specifically noting that the median start date for AI tool use was April 2024. The report underscores the importance of “tool‑calling” capabilities that allow models to fetch external information, compile code, and run tests, thereby improving self‑correction during development.
Salva’s Personal Workflow
Salva describes his own development routine as a mixture of hobby projects and professional responsibilities, all mediated through command‑line AI tools. He works with multiple IDEs—including Zed, VS Code, Cursor, and Windsurf—to read generated code, but relies on Gemini CLI to create detailed requirement documents in Markdown, generate code based on those specs, and iteratively update the documentation as the task progresses. Each AI‑driven change results in a separate commit and pull request, giving him the ability to revert or adjust as needed.
He estimates that roughly 70% to 80% of his work involves interacting with Gemini CLI via natural language, allowing the model to handle the bulk of coding while he focuses on review and architectural decisions.
Evolving Role of the IDE and the Developer
While the integrated development environment has long been the hub of software creation, Salva predicts its prominence will gradually decline as AI takes over routine coding tasks. Instead, developers will spend more time crafting high‑level requirements and guiding AI systems. In his view, the future developer will resemble an architect, breaking complex problems into manageable pieces and supervising AI‑generated solutions rather than manually writing each line of code.
Implications for the Industry
The shift toward AI‑assisted development suggests a long‑term transformation of the software engineering landscape. Salva acknowledges that the transition may span many years, but the current trajectory points toward a more collaborative relationship between humans and AI, where the latter handles repetitive coding work and the former focuses on strategic design and oversight.