Ryan Salva, Google’s AI coding tool lead, recently offered his perspective on the future of coding tools influenced by AI. His presentation was bolstered by the findings of a detailed report as well as his own lived experiences. Thanks to Salva’s team for putting together such an interesting new study and fueling this valuable discussion. It explored how developers adopted AI tools into their workflows post-April 2024.
Given those broader implications, the key research findings suggest that developers are already far ahead of the curve in how they utilize new AI tools. There is so much more progress to be made. Salva made the point that AI tool adoption is accelerating. Most developers continue to find it difficult to take full advantage of these technologies to increase their productivity.
Codi’s Code
Salva’s own coding practices echo a mix of different IDEs and command-line tools. He uses Cursor, Visual Studio Code and Zed as his main IDEs, suggesting an adaptable mindset towards coding. Most impressively of all, though, he uses command line based tools for virtually all his coding work, so it’s an even more efficient setup.
In his daily workflow, Salva uses Claude Code and Codex as well for more focused coding tasks. He encourages everyone to try using Gemini CLI to help outline their project needs. In fact, it even assists him in writing most of his own code. This method allows him to focus on reviewing and reading code within an IDE, ensuring quality and functionality.
About 70-80% of Salva’s labor is done in the terminal, and all of it is done in natural language. This philosophy is in step with the growing movement towards developer-first solutions. They are becoming more comfortable using AI tools to do their coding tasks. His involvement with product managers really brings to life how deeply AI has become ingrained into the development cycle. These managers often only interact with code through documentation, making AI-generated requirements and specification drafting the most useful tool.
“Here’s how we do testing, here’s how we manage dependencies, and so on.” – Ryan Salva
The shift to AI-assisted coding has sparked a transformation beyond technical workflows, by redefining the developer’s role and responsibilities. Salva, for his part, doesn’t have much doubt that developers will have to adjust their skill sets to man the controls of this new paradigm. He contends that development work in the future will look more like architecture work than coding work.
“I think that your job as a developer is going to look a lot more like an architect. It is going to be about taking big, complex problems and breaking them down into smaller, solvable tasks,” said Salva. “You’ll need to be thinking about the bigger picture about what you’re trying to produce, rather than the intermediate language in order to express that in machine code.”
Salva’s insights indicate a shift in how developers approach their work, necessitating a broader understanding of software architecture and problem-solving methodologies. As developers get comfortable with AI tools, they are reimagining what they want to do with their time. Now, creativity and strategic thinking trumps pure tech ability.
The standardized report from Salva’s team underscores the need for further research and development in AI coding tools. Pioneers are already adopting these assets as an essential part of their workflow. A major divide remains in getting the most out of them. Overall, the results point to the continued development of AI-based technologies in the future to help assist developers in producing efficient, high-quality code.