Do newer AI-native IDEs outperform other AI coding assistants?

​This post was originally published in Engineering Enablement, DX’s newsletter dedicated to sharing research and perspectives on developer productivity. Subscribe to be notified when we publish new issues.

For the second quarter in a row, DX data suggests that newer AI-native tools like Cursor and Claude Code outperform other AI coding assistants when measured by pull request merges per week.

Last week, DX published the AI-Assisted Engineering: Q4 2025 Impact Report, which shares AI usage and impact data from over 400 companies. One of the things we found: when looking at AI usage patterns by tool, we see that developers who use Cursor and Claude Code have a higher weekly PR merge rate than developers who use different tools, regardless of how regularly they use AI.

I was looking into why this might be, so I dug into the other factors that influence PR throughput.

An organization’s context—things like their tools, processes, and leadership—has a major influence on how much impact AI tools can actually make. It’s widely understood that bigger orgs are less efficient, while smaller and younger companies often operate with a faster rhythm of business. So what does that mean for AI tool performance?

Claude Code, Cursor, and GitHub Copilot are used by developers at all company sizes. Among the 200+ companies using Cursor in our dataset, 90% have fewer than 500 engineers. The remaining 10%—the larger enterprises—account for nearly half (46%) of all developers using Cursor.

The story is very different for Windsurf, which is taking a strong foothold in the enterprise. In our sample, nearly half of Windsurf’s end-user companies are large organizations, resulting in 65% more enterprise developers using Windsurf than Cursor. For Windsurf, this shows up as a lower PR throughput number. But this isn’t because of tool capabilities; larger orgs are less efficient, so they are starting with a lower PR throughput number.

I’m not completely surprised by how these adoption trends are taking shape. At the 2025 Enterprise Tech Leadership Summit in September, I noticed that Windsurf was popping up with disproportionate frequency in my conversations with enterprise tech leaders. It makes sense; most of these conversations were with leaders at enterprises who were early to the AI and got started with Codeium on-prem. Then, they looked toward Windsurf as an enterprise-ready agentic IDE to modernize their strategy. Our data largely supports my observation here, and this is a trend I’ll be keeping an eye on for the next few quarters.

Final thoughts

AI impact is influenced by existing organizational factors, and single metrics like PR throughout can only tell part of the story. Snapshots of industry trends like we shared in the AI-Assisted Engineering: Q4 2025 Impact Report can help contextualize your own performance, but at DX, we encourage you to measure comprehensively and continuously. Longitudinal data tells an important story of how your strategy is playing out over time. To add additional layers of insight, segmenting developers based on attributes – like seniority, tenure, tech stack, or even laptop age – enables you to perform cohort analysis at a level of granularity necessary to make meaningful decisions about your AI strategy.

While newer AI tools do show evidence of accelerating developers when it comes to PR merge frequency, there are many other factors to consider.