Windsurf CEO Farun Mohan at VB Transform 2025

With the spread of tools that work with AI materials through the chimneys of the institution’s programs, the rapid growth of the artificial intelligence coding platform Windsurf A case study of what is happening when developers adopt a widespread agent tools.

In a session per day VB Transform 2025and CEO and co -founder Farun Mohan Discuss how the integrated development environment in Windsurf (IDE) exceeded one million developers within four months of launch. More importantly, the basic system now writes more than half of the code committed by its user base.

The conversation, which was opened by the CEO of Venturebeat Matt Marshall, was opened with a brief but note: Mohan was unable to comment on the large -scale possible acquisition of Openai on Windsurf.

Case He drew attention after a Wall Street Journal a report In detail, a cheering confrontation between Openai and Microsoft regarding the terms of this deal and the broader tensions in its billion dollars in partnership. According to WsjOpenai seeks to obtain Windsurf without giving Microsoft to reach its intellectual ownership – an issue that can reshape the AI ​​coding scene of the Foundation.

With this allocated context, the session focused on Windsurf technology, traction of institutions, and the vision of the agent.

>>Watch every coverage of our conversion 2025 here<

Transgress

IDE Windsurf is built around what the company calls the “Mind” episode-a joint venture between humans and AI that allows full coding flows instead of additional automatic suggestions. Using this setting, agents can perform multi -shelves re -equipment, write test wings, and even launch user interface changes when starting a withdrawal request.

Mohan stressed that the coding assistance cannot stop when generating the code. “About 20 to 30 % of the developer time spends a code writing code. The rest is to correct errors, review and test. For help, the artificial intelligence system needs to reach all these data sources.”

Windsurf has recently included an IDE browser, allowing agents to test changes, read records and interact with live facades directly – as the human engineer does.

Security and control by design

When artificial intelligence begins to participate in institutional development courses actively, Windsurf’s concentration on integrated security has proven necessary. “We use a mixed model for the deployment of institutions – no personal data is stored outside the user’s tenant. Security is an essence, especially with features such as our integrated browser agent,” said Mohan.

These capabilities have already made Windsurf an applicable option for organized industries. Its agents are published on the huge code bases, including those in JPMorgan Chase and Morgan Stanley.

Mohan added that as artificial intelligence approaches roles, security will become a factor in productivity. He said: “If every person in a technology company contributes some way, then the lost piece is safety. You don’t want a single -time system created by a non -technical user who destroys another service.”

Small teams, fast test

Internally, Windsurf is organized in meager teams of three or four engineers, each focusing on testing a narrow range of product hypotheses.

“There is a belief that the companies of one billion dollars are the future. But in reality, more people allow you to grow faster and build better products. The key is organizing in small, concentrated teams that test the hypotheses in parallel,” Mohan explained.

This model helped the company quickly in an area where artificial artificial intelligence models – and user needs – quickly develop Breakneck.

Specialization on a large scale

The biggest improvement in Windsurf on Enterprise is not a generation of distinctive symbol or smaller models – it’s importance. “On a large scale, the greatest improvement is customization. Understanding a deep blade base allows the agent to make large -scale maintenance changes that reflect the user’s intention,” said Mohan.

Instead of relying only on the generation of the code for general purposes, the Windsurf system learns the structure of each customer, his style and preferences.

Building future models

Look forward, Windsurf is designing its platform to remain adaptable as the basic artificial intelligence models continue to grow.

Mohan said: “Every improved job in basic models requires rethinking the main products. While agents are more capable, our mission is to build the platform to manage many of them effectively.”

The company is working on an open protocol that allows institutions to integrate any LLM-including local models-in the Agent framework in Windsurf, maintain flexibility and reduce the seller lock.

Proof and measure value

Windsurf provides transparency in its performance with compact analyzes. Mohan said: “We offer transparency on the return on investment through measures – such as the percentage of the symbol written by the assistant – which can be attached directly to the internal engineering performance.”

This approach allows platform teams to link the product of the agent to the impact of business, which helps to justify more investment.

Concentrated implementation on flash

Finally, when asked how Windsurf plans to distinguish between the competition from Openai, Microsoft, Google, etc., Mohan focused on the interior speed. He said: “The challenge is not the most clear today, but who implements the correct strategy quickly enough. The risks are either moved very slowly or expect very far and lost importance in the short term.”

Mohan also rejected the idea that job holders are definitely governed. “There is no basic reason that old companies like Salesforce cannot become original. The real determinant is their speed in innovation, not their ability.”

Whether Windsurf becomes part of Openai’s future or continues independently, drag the company with institutional customers – and its insistence on grounding AI in the safe and valuable workforce – makes him a player worth seeing with his entry into developing the agent to the main current.