Cursor's coding agent has significantly impacted developer productivity, according to a recent study by Suproteem Sarkar from the University of Chicago. After the agent became the default tool for tens of thousands of users, companies experienced a 39% increase in merged pull requests (PRs). This suggests that the agent aids in streamlining the development process. Interestingly, experienced developers seem to engage more with the agent, writing more plans before coding and accepting agent-generated code at a higher rate compared to junior developers. For every additional year of experience, the acceptance of agent edits rises by roughly 6%. The study also examined user behavior regarding how often developers request agent assistance and their acceptance of its code edits. Junior developers tend to rely more on alternative methods, while senior developers show greater confidence in evaluating agent-produced code. This might reflect their ability to employ custom rules or manage context better, making them more adept at integrating the agent into their workflow. The research identified three main types of requests: implementing code, explaining code and errors, and planning actions, with implementation requests making up about 61% of interactions. In terms of overall productivity, while the increase in merged PRs is notable, some other metrics remained stable. The rate of PR reverts did not change significantly, and there was a slight decline in bug fixes. The average lines of code edited and files touched per merged PR also showed no significant variation. These findings indicate that while the agent enhances some aspects of productivity, the broader economic impact of AI in software engineering remains unclear. Further research is planned to better understand these effects.