The company's journey began as Exafunction, a GPU virtualization startup. When the market consolidated around the Transformer architecture and their autonomous vehicle customer base weakened, they executed a rapid and decisive pivot to become Windsurf, an AI developer tool.
The CEO advocates for balancing "irrational optimism" with "uncompromising realism." This philosophy drives a culture of rapid learning and development, with major product releases shipping every one to two weeks, based on the belief that speed of learning is a key competitive advantage.
Windsurf chose to build and deploy its own proprietary model, SWE1, rather than relying solely on third-party APIs. This model is optimized for their specific use case (agentic code workloads) and processes massive amounts of data daily, giving them control over performance, latency, and cost.
The vision is not to replace developers, but to augment them with AI agents that act as end-to-end assistants. This will elevate the level of abstraction, enabling more people (like technical PMs) to build software and allowing expert engineers to focus on more complex, mission-critical problems.
Despite being a cutting-edge AI company, a majority of Windsurf's revenue comes from enterprise customers. A key to this success is supporting the tools large organizations actually use, such as the Java language and JetBrains IDEs, not just the newer technologies popular in startups.
Keep pulling the thread on Varun Mohan.