Be a part of the occasion trusted by enterprise leaders for almost 20 years. VB Remodel brings collectively the folks constructing actual enterprise AI technique. Study extra
As AI-powered instruments unfold by means of enterprise software program stacks, the speedy progress of the AI coding platform Windsurf is turning into a case examine of what occurs when builders undertake agentic tooling at scale.
In a session at at this time’s VB Remodel 2025 convention, CEO and co-founder Varun Mohan mentioned how Windsurf’s built-in improvement atmosphere (IDE) surpassed a million builders inside 4 months of launch. Extra notably, the platform now writes over half of the code dedicated by its person base.
The dialog, moderated by VentureBeat CEO Matt Marshall, opened with a short however pointed disclaimer: Mohan couldn’t touch upon OpenAI’s extensively reported potential acquisition of Windsurf.
The problem has drawn consideration following a Wall Road Journal report detailing a brewing standoff between OpenAI and Microsoft over the phrases of that deal and broader tensions inside their multi-billion-dollar partnership. In keeping with the WSJ, OpenAI seeks to amass Windsurf with out giving Microsoft entry to its mental property— a difficulty that might reshape the enterprise AI coding panorama.
With that context put aside, the session centered on Windsurf’s expertise, enterprise traction, and imaginative and prescient for agentic improvement.
>>See all our Remodel 2025 protection right here<<Transferring previous autocomplete
Windsurf’s IDE is constructed round what the corporate calls a “mind-meld” loop—a shared undertaking state between people and AI that allows full coding flows somewhat than autocomplete strategies. With this setup, brokers can carry out multi-file refactors, write take a look at suites, and even launch UI modifications when a pull request is initiated.
Mohan emphasised that coding help can’t cease at code era. “Solely about 20 to 30% of a developer’s time is spent writing code. The remainder is debugging, reviewing, and testing. To actually help, an AI system wants entry to all these information sources,” he defined.
Windsurf lately embedded a browser inside its IDE, permitting brokers to check modifications, learn logs, and work together with reside interfaces immediately—very like a human engineer would.
Safety and management by design
As AI begins to take part in enterprise improvement cycles actively, Windsurf’s emphasis on built-in safety has confirmed important. “We use a hybrid mannequin for enterprise deployments—not one of the customized information is saved exterior the person’s tenant. Safety is core, particularly with options like our built-in browser agent,” mentioned Mohan.
These capabilities have already made Windsurf a viable possibility for regulated industries. Its brokers are deployed on huge codebases, together with these at JPMorgan Chase and Morgan Stanley.
Mohan added that as AI turns into extra accessible throughout roles, safety will develop into a gating issue for productiveness. “If everybody at an organization will contribute to expertise in a roundabout way, the lacking piece is safety. You don’t desire a one-off system constructed by a non-technical person destroying one other service,” he mentioned.
Small groups, speedy testing
Internally, Windsurf organizes into lean squads of three or 4 engineers, every centered on testing a slim set of product hypotheses.
“There’s a perception that one-person billion-dollar corporations are the longer term. However in actuality, extra folks permit you to develop sooner and construct higher merchandise. The bottom line is organizing into small, centered groups that take a look at hypotheses in parallel,” Mohan defined.
This mannequin has helped the corporate iterate rapidly in an area the place foundational AI fashions—and person wants—evolve at breakneck pace.
Personalization at scale
Windsurf’s largest optimization at enterprise scale isn’t sooner token era or smaller fashions—it’s relevance. “At scale, the most important optimization is personalization. Deeply understanding the codebase permits the agent to make maintainable, large-scale modifications that mirror person intent,” mentioned Mohan.
Moderately than relying solely on general-purpose code era, Windsurf’s system learns every buyer’s stack’s construction, type, and preferences.
Constructing for future fashions
Wanting ahead, Windsurf is designing its platform to stay adaptable because the capabilities of underlying AI fashions proceed to develop.
“Each step-function enchancment in basis fashions calls for a serious product rethink. As brokers develop into extra succesful, our job is to construct the platform to handle and orchestrate lots of them successfully,” Mohan mentioned.
The corporate is engaged on an open protocol that can enable enterprises to combine any LLM—together with on-premises fashions—into Windsurf’s agent framework, preserving flexibility and minimizing vendor lock-in.
Proving and measuring worth
Windsurf supplies transparency into its efficiency with built-in analytics. “We offer transparency on ROI by means of metrics—like the proportion of code written by the assistant—which could be tied on to inner engineering efficiency,” Mohan mentioned.
This method permits platform groups to attach agentic productiveness with enterprise impression, serving to justify additional funding.
Centered execution over flash
Lastly, when requested how Windsurf plans to distinguish itself amid competitors from OpenAI, Microsoft, Google, and others, Mohan centered on inner velocity. “The problem isn’t who’s extra seen at this time, however who’s executing the fitting technique quick sufficient. The danger is both transferring too slowly or projecting too far out and lacking near-term relevance,” he mentioned.
Mohan additionally dismissed the concept incumbents are inevitably doomed. “There’s no basic cause legacy corporations like Salesforce can’t develop into AI-native. The actual limiter is their pace of innovation, not their functionality.”
Whether or not Windsurf turns into a part of OpenAI’s future or continues independently, the corporate’s traction with enterprise clients — and its insistence on grounding AI in safe, measurable workflows — makes it a participant value watching as agentic improvement enters the mainstream.
