Building a magical AI code editor used by over 1 million developers in four months: The untold story of Windsurf | Varun Mohan (co-founder and CEO)

Apr 20, 2025 Episode Page ↗
Overview

Varun Mohan, co-founder and CEO of Windsurf, discusses their AI coding tool's rapid growth, strategic pivots, and the importance of "agency" for engineers. He shares insights on the future of coding, unique hiring, and product development.

At a Glance
27 Insights
1h 14m Duration
18 Topics
6 Concepts

Deep Dive Analysis

Varun Mohan's Background and Company Pivots

From GPU Infrastructure to AI Coding Tool

The Genesis and Purpose of Windsurf IDE

UI Discovery Triples Adoption Rates

Windsurf's Traction and Growth

The Future of Engineering and AI's Role

Essential Skills for the AI Era: Agency and Problem Solving

Windsurf's Lean Hiring Philosophy and Culture

Strategic Investment in Enterprise Sales Early On

Differentiating Windsurf from Competitors

Live Demo: Building an Airbnb for Dogs with Windsurf

Tips for Effective Use of AI Coding Tools

AI's Role in Code Modification and Review

Empowering Non-Developers to Build Custom Software

Windsurf's Model Training and Data Advantage

Company Structure and Product Strategy

The Importance of Continuous Innovation and Long-Term Bets

Advice for Aspiring Developers in the AI Era

Startup Pivoting

Startups must be willing to abandon working ideas when core assumptions change, focusing on the underlying problem rather than being overly attached to a specific solution. This requires a balance of irrational optimism and uncompromising realism about market shifts.

Agency (in AI Era)

Agency is the undervalued skill of taking initiative and building something, rather than merely following prescribed paths. In the AI era, it becomes crucial for individuals to identify problems and proactively create solutions, leveraging AI tools to multiply their impact.

Amdahl's Law

This principle from parallel computing states that the overall speedup of a program by parallelizing a task is limited by the time spent on the non-parallelizable, sequential portion. It illustrates that even if AI automates 90% of code writing, the total engineering process won't see a 90% speedup because other tasks like debugging, testing, and design still exist.

Cannibalizing Your Product

A strategic approach where a company intentionally develops new products or features that make its existing offerings seem outdated or 'silly' every 6-12 months. This ensures continuous innovation, prevents stagnation, and maintains market leadership by proactively disrupting one's own products.

Dehydrated Entity (Hiring)

A metaphor for a company's hiring philosophy where new hires are only brought on when existing teams are 'underwater' and desperately need more capacity. This approach forces ruthless prioritization of tasks and prevents internal politics that can arise from having excess staff.

Code Base Understanding

The capability of AI models to deeply analyze and comprehend very large, complex code repositories, sometimes exceeding 100 million lines of code. This understanding is crucial for making large-scale changes, providing relevant context for development, and enabling agents to operate effectively within existing projects.

?
How will the role of engineers change with AI writing most of the code?

Engineers will shift their focus from 'solving it' (pure code writing) to 'what should I solve for?' and 'how should I solve it,' prioritizing important business problems and making high-level technical and design decisions.

?
Should aspiring students still pursue a computer science degree in the AI era?

Yes, a computer science degree remains valuable as it teaches fundamental problem-solving principles and mental models of how computers and systems work, which are crucial for making informed design and architectural decisions, even if AI writes much of the code.

?
What is Windsurf's philosophy on hiring and team size?

Windsurf aims to be the smallest company possible to achieve its ambitions, hiring only when existing teams are 'underwater' and desperately need more capacity, which forces ruthless prioritization and prevents internal politics.

?
Why did Windsurf invest early in a large enterprise sales team?

Windsurf quickly identified a significant enterprise market for its product, realizing that large companies like Fortune 500 cannot be served purely through product-led growth and require a dedicated sales motion.

?
How does Windsurf differentiate itself from competitors like Cursor?

Windsurf focuses on high-quality code base understanding for very large codebases, supports multiple IDE platforms (like JetBrains), and handles complex secure enterprise environments with FedRAMP compliance and hybrid deployment modes.

?
Can non-technical staff use Windsurf to build custom software?

Yes, Windsurf empowers non-developers to build custom applications; for example, its go-to-market team built internal tools, saving over $500,000 in SaaS product costs.

?
How does Windsurf train its AI models to gain an edge?

Windsurf leverages millions of pieces of user feedback, including preference data from users typing incomplete code, allowing it to build models uniquely capable of completing code in intermediate states, unlike frontier models trained on complete code.

?
What is the most counterintuitive lesson learned about building AI products?

The most counterintuitive lesson is the need for continuous, self-disruptive innovation, where the company aims to 'cannibalize' its existing product every 6-12 months with new, unasked-for features, rather than just incremental improvements.

?
What is the most important skill for individuals to invest in for the AI era?

Agency, the drive to build and take initiative, is crucial, as is getting hands-on experience with AI tools to understand their capabilities and limitations, becoming a force multiplier in any organization.

1. Get Hands-On with AI Tools

Actively engage with and learn to use new AI products and tools as quickly as possible. This allows you to become a “force multiplier” in your organization, gaining a significant productivity advantage.

2. Continuously Cannibalize Your Product

Aim to make your existing product look “silly” or “dumb” every 6-12 months by developing new, disruptive innovations. This proactive disruption ensures long-term relevance and prevents stagnation.

3. Hire Only When Dehydrated

Only add new team members when the existing team is “underwater” and truly needs support. This forces ruthless prioritization, prevents internal politics, and ensures every hire is impactful.

4. Ruthless Prioritization for Focus

Concentrate efforts on doing one thing exceptionally well, even if it means deprioritizing or failing at others. This focused approach is critical for startup success.

5. Cultivate High Agency

Develop and prioritize “agency”—the drive and initiative to build things and solve problems independently, rather than just following prescribed paths. This skill is crucial for innovation and adapting to rapid change.

6. Empower Domain Specialists to Build Custom Tools

Enable non-engineering domain specialists (e.g., sales, finance) to build their own custom software using AI. This eliminates reliance on “kitchen sink” SaaS products and creates highly tailored solutions.

7. Product Managers Learn to Code with AI

Product managers should learn to use AI coding tools to make direct edits and push changes themselves. This increases their output, earns respect from engineering, and removes traditional role barriers.

8. Balance Optimism with Realism

Be irrationally optimistic about your vision but also ruthlessly realistic, constantly testing hypotheses. Be willing to “kill” beliefs that are proven wrong by new information.

9. Maintain Laser Focus

When pursuing a new, big idea, fully commit and pivot away from existing efforts you no longer believe are valuable. Divided focus guarantees failure for ambitious endeavors.

10. Prioritize Truth Over Ideas

Avoid becoming overly attached to your own ideas; instead, foster an organizational culture that is “truth-seeking.” This allows for constant testing and adaptation without fear of being wrong.

11. Be Patient and Explicit with AI Tools

When using AI coding tools, be patient and as explicit as possible with your requests, starting with smaller changes. This helps prevent irrelevant outputs and allows you to learn the tool’s capabilities.

12. Leverage AI for Code Refactoring

When refactoring or migrating code, make an initial change yourself and then instruct the AI to propagate that change across the codebase. The AI’s deep understanding allows it to find and modify all corresponding locations.

13. Understand AI Tool Capabilities

Learn the strengths and weaknesses (“hills and valleys”) of AI tools through active use. This allows you to effectively leverage them where they excel and anticipate their limitations.

14. Focus Engineers on Business Problems

Shift engineering focus from pure code writing to identifying and prioritizing the most important business problems and product capabilities. AI handles the “solving it” part, freeing engineers for strategic decisions.

15. Build Foundational System Knowledge

Invest in understanding the underlying principles of how computers and systems work (e.g., operating systems, parallel computing). This foundational knowledge helps make better design decisions and diagnose performance issues.

16. Optimize for Lean Ambition

Strive to be the smallest possible company that can still achieve its ambitious goals. This ensures resources are focused on high-impact work rather than idolizing leanness for its own sake.

17. Avoid Over-Hiring to Prevent Politics

Resist hiring for roles where current staff are not “underwater,” as over-hiring leads to manufactured work and internal politics. This diverts focus from truly important tasks.

18. Value Project Ownership, Not Team Size

Structure teams around projects with directly responsible individuals, allowing flexible movement between important projects. This avoids “owning people” and rewards impact with minimal resources.

19. Hire for Passion and Hard Work

Beyond technical skills, seek candidates who are deeply passionate about the mission and explicitly willing to work very hard. This builds a collaborative culture with high standards.

20. Maintain High Work Ethic Bar

Set a high expectation for effort and commitment across the team. A single team member not pulling their weight can lower morale and the perceived standard for the entire group.

21. Embrace AI Tools in Interviews

Allow candidates to use AI coding tools during interviews, as these are seen as massive productivity improvements. Assess problem-solving ability and how they leverage tools, not just raw coding.

22. Invest in Enterprise Sales Early

Recognize the value of enterprise sales, especially when targeting large companies like the Fortune 500. Product-led growth alone is often insufficient for this market segment.

23. Directly Edit UI Elements with AI

Utilize AI coding tools like Windsurf to directly select and modify specific UI elements on a live preview. This allows for design changes without needing to write code manually.

24. Empower Engineers as Product Managers (for Dev Tools)

For developer-focused products, empower core engineers to act as product managers. This leverages their deep intuition and understanding of the user persona for streamlined decision-making.

25. Keep Teams Small for Technical Depth

Maintain small “two-pizza” team sizes to ensure leaders remain deeply involved and knowledgeable about the technology. This prevents “armchair quarterbacking” in fast-moving technical spaces.

26. Increase Investment in Engineering with AI

Recognize that AI increases the ROI of building technology, making the opportunity cost of not investing higher. Companies should hire more engineers to leverage enhanced productivity.

27. Embrace Discomfort to Act Faster

Cultivate the ability to reevaluate hypotheses and make difficult decisions faster, even if it means entering “uncomfortable” territory. Proactive adaptation is key to staying ahead.

Every six to 12 months, it should make our existing product look silly. It should almost make the form factor of existing product look dumb.

Varun Mohan

I wanted the company to almost be like this dehydrated entity. Every hire is like a little bit of water. And we only go back and hire someone when we're back to being dehydrated.

Varun Mohan

You don't win by doing, you know, 10 things kind of well, you win by doing like one thing really well, and maybe you fail nine things.

Varun Mohan

If we don't innovate and do crazy things, we're going to die, the company is just going to die.

Varun Mohan

The domain specialists now have access to build the tools that they ultimately wanted.

Varun Mohan

If the only way you can solve a hard problem is like put it into chat to PT. I think, I think that's like a concern to us.

Varun Mohan

The ROI of building technology has actually gone up. So the opportunity cost of not investing more into technology has gone up, which means that you should just invest even more.

Varun Mohan
Over 1 million
Windsurf users Developers who have tried the product in a bit over four months since launch.
Many hundreds of thousands
Windsurf monthly active users Current monthly active users.
90%
AI code generation prediction Predicted percentage of code that will be AI-generated in the future.
Over 80 people
Windsurf sales team size Current size of the go-to-market team.
Close to 160 people
Codium total employees Total company headcount.
Over 50 people
Codium engineering team size Size of the core engineering team.
70-80%
JetBrains market share (Java developers) Percentage of Java developers who code in JetBrains-based IDEs like IntelliJ.
Over $500,000
SaaS cost savings (internal tools) Amount saved by Windsurf's non-developer teams building custom apps instead of purchasing SaaS products.
$17 billion
JP Morgan Chase annual software budget CIO's budget for software every year at a company like JP Morgan Chase.
Over 50,000 engineers
JP Morgan Chase engineers Number of engineers inside a company like JP Morgan Chase.
Over 100 million lines of code
Code base size (Dell example) Size of singular code bases at large companies like Dell.
Over 30%, close to 40%
AI productivity improvements for engineers Observed productivity improvements for engineers using AI tools.
0.6%
Engineering interview acceptance rate (post-take-home) Acceptance rate for engineering candidates at Codium after the take-home assignment.
99%
Time to build apps/technology reduction goal Windsurf's ambitious goal for reducing the time it takes to build applications and technology.