Cursor's most important AI features started as side projects engineers built themselves, says its engineering head

In a recent episode of the "LangChain" podcast, Jason Ginsberg, Cursor's engineering head, revealed that some of the company's most significant AI features emerged from informal, grassroots initiatives rather than structured plans. Ginsberg shared a notable example where he developed a debugging tool over the Thanksgiving holiday, driven by personal necessity and a desire to support his colleagues. This initiative eventually led to the launch of Cursor's 'Debug Mode.' Ginsberg emphasized that internal adoption serves as a key indicator for determining when a feature is ready for broader release. This bottom-up approach has also been instrumental in the development of Cursor's agent feature, which began as a prototype created by a single engineer amidst initial skepticism from the team. The surprising success of the prototype prompted a shift in perception, with Ginsberg stating, "Oh wow, this works." While Cursor maintains short-term roadmaps, Ginsberg noted that many of the company's most impactful features arise organically. The company fosters a culture where engineers resolve disputes through coding rather than lengthy discussions in meetings or documentation. With a compact team of around 20 employees as of early 2025, Cursor exemplifies the trend of small, highly skilled teams driving innovation in the AI sector. Ginsberg attributed this structure to slow hiring processes and high standards for new talent, which enables Cursor to operate efficiently with minimal formal processes. This preference for elite teams is gaining traction across the AI landscape, even among larger corporations like Meta, which has a small, specialized unit focused on superintelligent AI. Mark Zuckerberg has expressed his belief in the effectiveness of talent-dense teams for advancing cutting-edge research, reflecting a broader trend in the industry. OpenAI's CEO, Sam Altman, also predicts that we will soon see small startups achieving billion-dollar valuations. In May, Business Insider highlighted a list of highly valued AI startups with teams of 50 or fewer, showcasing this emerging dynamic within the tech landscape.