Cursor’s key AI capabilities were built from the bottom up, says engineering lead

AI For Business


Some of Cursor's most important AI features didn't come from a formal roadmap. One of them was built during Thanksgiving, the engineering director said.

Jason Ginsberg said on Thursday's episode of the LangChain podcast that the bottom-up approach formed some of Cursor's core features.

Ginsburg said he built the debugging feature over the Thanksgiving holiday simply because he wanted it and “to help people on the team.” The AI ​​coding company then entered “debug mode.”

“If we have internal adoption, that's an indicator that we're ready to ship,” he said.

The same pattern applies to Cursor's agents and is now one of its characteristics. Ginsburg said the prototype was originally built by one engineer because the rest of the team was skeptical.

“He prototyped it so quickly that everyone was like, 'Oh, this works,'” Ginsburg said.

While Cursor still maintains a short-term roadmap, many of its biggest features have emerged organically, Ginsburg said.

He also said Cursor doesn't have much of a formal process. Engineers resolve disagreements through code rather than discussing product changes in documents or coordination meetings.

Small team, light process

Cursor is one of the most prominent AI companies built by a small team.

Ginsburg said on the podcast that as of early 2025, Cursor had about 20 employees.

“That's because the hiring process was very slow and the hurdles were very high,” he says.

This talent-dense structure allows Cursor to operate with minimal organizational processes and move quickly, he added.

The preference for small, elite teams is becoming increasingly influential across the AI ​​industry, including among Big Tech companies, which are traditionally known for their size.

For example, Meta's Superintelligence AI division is led by a small group of top researchers. The AI ​​division makes up a small portion of the company's workforce of more than 70,000 people.

“We're a little more confident that a small, talented team is the best structure to drive frontier research,” Mark Zuckerberg said during Meta's July earnings call.

OpenAI CEO Sam Altman said last year that “very soon there will be 10-person companies valued at $1 billion.”

In May, Business Insider compiled a list of the highest-rated AI startups around the world with teams of 50 or fewer employees, according to PitchBook data.





Source link