Cursor’s Top AI Features Were Built Bottom-up, Says Engineering Head

Some of Cursor’s most important AI features didn’t come from a formal road map. One was built over Thanksgiving, says its engineering head.

Jason Ginsberg said in an episode of the “LangChain” podcast published Thursday that a bottom-up approach has shaped several of Cursor’s core capabilities.

Ginsberg said he built a debugging feature over the Thanksgiving holiday simply because he wanted it and to “help people on the team.” The AI-coding company has since launched its “Debug Mode.”

“If there’s internal adoption, that’s kind of our metric for this is ready to ship,” he said.

The same pattern applies to Cursor’s agent, now one of its defining features. Ginsberg said it was originally prototyped by a single engineer, as others on the team were skeptical.

“He prototyped it super quickly, and everyone’s like, ‘Oh wow, this works,'” Ginsberg said.

Cursor still maintains short-term roadmaps, but many of its biggest features emerge organically, Ginsberg said.

He also said there isn’t much formal process at Cursor. Instead of debating product changes in documents or alignment meetings, engineers resolve disagreements through code.

Small teams, light processes

Cursor is among the most prominent AI companies being built by a tiny team.

Ginsberg said on the podcast that Cursor had about 20 people at the start of 2025.

“That was because the process to hiring people was very slow and the bar was extremely, extremely high,” he said.

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

The preference for small, elite teams has become increasingly influential across the AI industry, including at Big Tech firms traditionally known for scale.

Meta’s superintelligence AI unit, for example, is led by a small group of top researchers. The AI unit comprises a tiny fraction of the company’s headcount of over 70,000 employees.

“I’ve just gotten a little bit more convinced around the ability for small, talent-dense teams to be the optimal configuration for driving frontier research,” Mark Zuckerberg said on Meta’s earnings call in July.

OpenAI CEO Sam Altman said last year that “we’re going to see 10-person companies with billion-dollar valuations pretty soon.”

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