The Cursor Playbook: How Four MIT Graduates Built the Fastest-Growing SaaS Company in History

On Tuesday I told you about Cristóbal Valenzuela, the designer from Santiago who moved to New York with no AI background, got rejected by 50 investors, built Runway, and ended up with every major film studio as a customer and Adobe Ventures on his cap table.

Go read that one if you missed it.

Today's teardown is about a completely different company and a completely different founding story. But there's one thing Runway and Cursor have in common:

Both were built by people who weren't supposed to win.

Runway was built by artists competing against Adobe and Hollywood.

Cursor was built by four MIT students competing against Microsoft.

And both of them won.

📌 No sponsor this week - Projects of AI just relaunched. If your product belongs in front of 2,400+ AI founders and builders with a 69% open rate, grab a slot here →

The founding story that almost didn't happen

Four MIT classmates launched Anysphere in 2022: Michael Truell (CEO), Sualeh Asif (CPO), Arvid Lunnemark, and Aman Sanger.

They met studying computer science and mathematics at MIT. They were obsessed with AI before most people knew what an LLM was. And they were watching GitHub Copilot, Microsoft's AI coding assistant and feeling frustrated.

Not because Copilot was bad. Because it wasn't pushing hard enough.

Copilot autocompleted. It suggested the next line. It was useful the way a spell-checker is useful, it caught mistakes and saved keystrokes. But it didn't understand your codebase. It didn't know what you were trying to build. It wasn't a collaborator. It was autocomplete on steroids.

The four co-founders believed something different was possible: an editor where AI didn't just assist your coding, it understood your entire project, your intentions, your architecture, and worked with you the way a senior engineer would.

In 2022, fresh out of MIT, they incorporated Anysphere, turning down job offers from Google, Jane Street, Stripe, and other institutions that compete aggressively for MIT CS graduates.

But here's the detail most people miss: they didn't immediately build Cursor. They spent nearly a year working on mechanical engineering tools first. A completely different market. It didn't work. So they pivoted.

The company that became the fastest-growing SaaS in history started with a year of building something nobody wanted. The thing that worked came after the thing that didn't.

Lunnemark departed in October 2025 to found his own company, Integrous Research leaving Truell, Asif, and Sanger to lead Cursor through its most explosive growth phase.

The one product decision that defined everything

In 2023, the team faced a choice that would define what Cursor became.

They could build Cursor as a plugin, a GitHub Copilot-style extension sitting inside VS Code. Quick to ship. Easy to distribute. No friction for developers already in their workflow.

Or they could build a standalone editor, a complete development environment built from scratch around AI, with intelligence at the center rather than bolted on.

They chose the standalone editor. The reasoning: if AI is genuinely going to change how developers work, not just autocomplete, but understand codebases, plan implementations, debug across files then the architecture of the editor itself needs to be redesigned around that capability. You can't bolt intelligence onto a tool designed before intelligence existed.

This decision made Cursor harder to build and slower to ship. It also made it impossible to replicate quickly once it worked.

The "Tab Tab Tab" experience where Cursor predicts exactly what you want to type, across multiple lines, understanding context from files you haven't even opened only works because the editor was designed around the model, not the model designed around the editor.

Design the product around the capability. Not the capability around the product.

The growth numbers that broke the SaaS record book

By early June 2026, Cursor's annualised revenue had climbed to $4 billion, more than doubling from the $2 billion run rate the company posted just four months earlier in February.

Let me put that trajectory in context because it's genuinely without precedent in software history:

• October 2023: $8M seed round. OpenAI Startup Fund led. Nat Friedman and Arash Ferdowsi as angels.

• Early 2024: $100M ARR milestone.

• November 2024: $400M Series B at $9.9B valuation. Thrive Capital led.

• November 2025: $500M ARR. $2.3B Series D. $29.3B valuation.

• February 2026: $2B ARR run rate.

• Early June 2026: $4B ARR - doubled in four months.

• June 16, 2026: SpaceX acquires Anysphere for $60 billion.

From $100M to $4B ARR in under two and a half years.

The SaaS comparison nobody talks about enough:

• Salesforce — 10 years to reach $1B ARR

• Slack — 7 years

• Zoom — 9 years

• Cursor — under 2 years

That's not a growth story. That's something different entirely.

How they grew without spending a dollar on marketing

Cursor reached $200M ARR with $0 spent on traditional marketing.

No paid ads. No growth team. No outbound sales for individual users.

The entire growth engine was word of mouth among developers, the hardest audience on earth to acquire artificially, and the most powerful audience on earth to acquire organically.

Here's why it worked.

Developers don't share tools unless those tools make them feel genuinely, uncomfortably more capable. The bar for a developer recommendation is "this made me feel like I was cheating", not "this saved me 10%."

Cursor cleared that bar on the first day for most users. The Tab Tab Tab experience where the editor predicts multi-line implementations correctly, understanding what you're building from context you haven't explicitly provided is the kind of experience that makes developers immediately open Slack and say: "you have to try this right now."

Every developer who experienced it became a distributor. Every team that adopted it became a reference customer for the next team.

Cursor was ranked No. 37 on the CNBC Disruptor 50 list in 2026 used internally by developers at NVIDIA, Adobe, Uber, Shopify, and PayPal. Wh en engineers at those organisations recommend a tool, it carries institutional credibility that no marketing budget can manufacture.

That's the flywheel: product experience → developer word of mouth → credibility via prestigious users → more adoption → more word of mouth.

Three companies tried to buy them. All three failed.

This is the part of the Cursor story that tells you everything about where the founders thought they were going.

In 2024, OpenAI made its first acquisition offer to Anysphere, which was rejected. In early 2025, OpenAI returned with another offer, which was again turned down. At the time, Anysphere believed its independent growth prospects were excellent and had no intention to sell.

Microsoft also explored acquiring Cursor but ultimately decided not to proceed with a bid due to concerns about regulatory scrutiny.

After failing to acquire Cursor twice, OpenAI shifted focus to Windsurf, another AI coding startup. Those talks ultimately collapsed over the team's concerns that their technology might indirectly flow to Microsoft. Windsurf's core team was later poached en masse by Google.

So the sequence runs: OpenAI tried twice and got rejected. Microsoft looked and walked away. OpenAI went to Windsurf instead and that deal collapsed too, with the team ending up at Google.

And then in April 2026, SpaceX secured an option to acquire Anysphere for $60 billion with a $10 billion break-up fee if the deal fell through.

On June 16, 2026, SpaceX confirmed the acquisition. Cursor will become a wholly owned subsidiary upon closing in Q3 2026. The deal absorbs Cursor into SpaceX's xAI ecosystem to supercharge Grok and scale enterprise software automation using the Colossus supercomputer cluster.

The company that turned down OpenAI twice, sold to SpaceX for $60 billion. At $60 billion, SpaceX paid approximately 15x Cursor's ARR, one of the largest multiples ever paid for an AI software company.

The company that said no to OpenAI ended up selling for twenty times what OpenAI eventually paid for their next-best option.

Three things any builder can steal right now

1. Design the product around the capability, not the capability around the product.

Cursor's standalone editor decision building an IDE from scratch for AI rather than a plugin for existing IDEs is the reason the experience feels different from every competitor. Every rival who chose the plugin route made a faster short-term decision and a worse long-term one. Whatever you're building: if your core capability genuinely changes what's possible, design the product around it from day one. Retrofitting never works as well.

2. Target the audience that talks most.

Developers are nearly impossible to reach through advertising. They're also the most powerful word-of-mouth network on earth. Cursor didn't market to developers, it made developers feel compelled to tell other developers. Find the audience in your space who, if they love your product, will tell everyone they know immediately. That audience is worth more than any reach you can buy.

3. Know what your independence is worth and don't sell for less.

Cursor said no to OpenAI in 2024. Said no again in early 2025. Kept building. Kept doubling. By June 2026, SpaceX paid $60 billion about twenty times what OpenAI paid for Windsurf, their second choice.

Every founder gets acquisition interest before they're ready to sell. The Cursor lesson isn't "never sell." It's: know what your independence is worth, and don't accept less. The two rejections of OpenAI didn't just preserve Cursor's upside, they multiplied it by twenty.

The ending nobody saw coming

In December 2025, Truell said publicly the company had no plans to IPO soon.

Six months later, SpaceX announced the $60 billion deal.

"We are excited to share that SpaceX has exercised their option to acquire Cursor in an all-stock transaction with the goal of building the world's most useful AI models," Truell said.

SpaceX's market cap topped Amazon and Microsoft following the announcement becoming the fourth most valuable company in the US.

Four MIT graduates. Turned down lucrative job offers after graduation. Spent a year on a product nobody wanted. Pivoted. Built the fastest-growing SaaS company in history. Turned down OpenAI twice. Turned down Microsoft. Got acquired by SpaceX for $60 billion.

The company that couldn't get traction in 2022 sold for more than the GDP of Luxembourg four years later.

That's Month 1 of Projects of AI - 8 issues, 4 companies, every story told the way it actually happened.

Next Tuesday we start Month 2. I'm covering Anthropic - the safety lab that left OpenAI, raised billions on the promise of building AI responsibly, and became OpenAI's most credible rival. There's more drama in that story than any we've covered so far.

Before then: hit reply and tell me which of these 8 stories hit hardest. I read every reply and use the answers to plan what comes next.

- Projects of AI

That's the Cursor story. The full 8-issue archive Midjourney, ElevenLabs, Perplexity, Runway, Cursor is at projectsofai.beehiiv.com. Free, twice a week. Forward this to one developer who'd find it useful.

P.S. - The sequence I keep coming back to: OpenAI approached Cursor in 2024 - rejected. Came back in early 2025 rejected again. Went to Windsurf instead that deal collapsed and the team ended up at Google. Meanwhile Cursor kept doubling. SpaceX paid $60 billion. That's three companies who wanted Cursor and couldn't get it and one that paid $60B when the moment was right. Saying no, twice, was the most valuable decision the Cursor team ever made.

Enjoying this? Forward it to one person building in AI. That's the only growth strategy I have.

Want to reach 2400+ AI founders and builders? Sponsor Projects of AI