Google’s Project Manager Ryan Salva Explores the Future of AI-Powered Coding and Its Impact on Developer Roles
Google’s project manager for developer tools, Ryan Salva, offers insights into the transformative impact of AI on coding. With a background at GitHub and Microsoft, Salva now leads teams responsible for innovative tools such as Gemini CLI and Gemini Code Assist, pushing developers towards the future of agentic programming.
In a recently released third-party research report, Salva’s team revealed valuable insights into how developers currently utilize AI tools and the progress that still needs to be made in this area. In an interview, we discussed the findings and Salva’s personal experience with AI coding tools.
Each year, Google conducts a survey on developer trends, but this year’s report focuses significantly on AI tools and their growing influence on programming methodologies. One intriguing revelation from the study was the median date developers started using AI tools, which was discovered to be April 2024, aligning with the release of Claude 3 and Gemini 2.5. This period marks a significant milestone in the development of reasoning or thinking models, along with improvements in tool-calling capabilities.
For coding tasks, Salva emphasizes the importance of leveraging external information to facilitate problem-solving. This may involve grep searches, code compilations, and running unit tests or integration tests. He believes that tool-calling is a crucial component that allows models to self-correct as they progress through tasks.
When it comes to personal usage, Salva primarily codes for hobby projects using command line-based tools like Gemini CLI, with minor integrations of Claude Code and Codex. His varied IDE choices include Zed, VS code, Cursor, and Windsurf, as he explores industry advancements across different platforms.
In a professional context, product managers predominantly work within documents. Salva uses AI to assist in writing specification and requirements documents. For him, the process involves utilizing Gemini CLI to create more detailed requirement docs in Markdown format, which then serve as a foundation for coding tasks using the specified requirements and team guidelines.
Salva’s engineering team follows multiple layers of rules and Markdown docs that are consumed by the model during the development process. As the tool troubleshoots issues, it updates the requirement documents to reflect each completed step, creating commits and pull requests within the repository for easy tracking and reversion if necessary.
Approximately 70-80% of Salva’s work involves him working in the terminal with natural language, using Gemini CLI to draft requirements, while allowing it to write most of the code. He then reviews and reads the generated code using whichever IDE he chooses for reading purposes.
Regarding the future of raw computer code, Salva believes that the role of IDEs may gradually diminish as developers spend more time working with requirements. While there is uncertainty about what this means for software development professionals, Salva envisions a future where the job of a developer becomes more architect-like, focusing on breaking down complex problems into manageable tasks and considering the bigger picture rather than the intermediate machine code.
