Impact of coding agents on software development velocity

Software developer Simon Willison recently analyzed his own coding output to measure the impact of advanced AI coding agents on his productivity. By examining the code-frequency chart for his open-source project, Datasette, he identified a significant surge in activity that correlates directly with the release of powerful new models, including Opus 4.8, GPT-5.5, and GPT-5.6.

This data suggests that modern AI tools are substantially increasing the speed at which developers can write and update software. For instance, Willison noted that a recent version of his sqlite-utils project was written almost entirely with the help of an AI model for a relatively low cost. The visual evidence from his GitHub repository shows a clear upward trend in code commits following the integration of these sophisticated generative tools into his workflow.

These findings highlight a shift in software development where AI agents act as force multipliers for individual programmers. By automating large portions of the coding process, these models allow developers to maintain higher levels of output and accelerate the release of new features and project updates. This trend indicates that the adoption of agentic AI is fundamentally changing the velocity and efficiency of open-source software maintenance.