Cursor
codes itself
into history

The room where it happens is a terminal window

In the spring of 2026, somewhere inside Nvidia's Santa Clara campus, 40,000 engineers open the same application to start their day. It is not Outlook. It is not Slack. It is Cursor - an AI code editor built by a 300-person startup in San Francisco that fewer people could name than the IDE itself.

Cursor is the product of Anysphere, a company founded in 2022 by four MIT graduates who looked at the software development stack and made a bet that most of their peers considered reckless: that AI would not merely assist programmers, but fundamentally replace the act of line-by-line typing altogether. Three years later, with $2 billion in annual recurring revenue and 70% of the Fortune 1000 as paying customers, the argument seems less reckless and more simply correct.

What makes Cursor distinct is not any single feature. It is the ambition of the product thesis: that the IDE is not the end product, but the starting point. With Cursor 3 - released April 2026 and built from scratch under the internal codename Glass - the company has stopped pretending the editor is the main event. The main event is the agents. The IDE is just where you watch them work.

Software development had a dirty secret: most of it is typing

The traditional IDE - a text editor with autocomplete and a debugger bolted on - had not fundamentally changed since the 1990s. Tools like VS Code improved the aesthetics and added extensions, but the underlying workflow remained the same: a human types code, looks up documentation, reads error messages, and types more code. The feedback loop runs on human time.

By 2022, large language models were producing code that could pass junior developer interviews. Yet every major IDE treated this new capability as an optional plugin - an extra menu item named "AI Assistant" that you might click on Tuesdays. The integration was cosmetic. The workflow was unchanged.

This is the gap Cursor targeted. Not AI as a feature, but AI as the primary interface. The cursor in a text editor, it turns out, is not just a blinking line indicating where you type next. It is a metaphor - one the founders decided to make literal.

Four MIT graduates who turned down everything sensible

Michael Truell (CEO), Sualeh Asif (CPO), Aman Sanger (COO), and Arvid Lunnemark (CTO, departed October 2025) were computer science and mathematics students at MIT who met through the Neo Scholars program - a competitive track for exceptional startup talent. Each had competed in the International Mathematical Olympiad. Each had internship offers from the companies everyone at MIT fought to join.

They rejected those offers. They incorporated Anysphere in 2022 with the conviction that the AI revolution in software was not happening fast enough and that existing tools were failing to push its limits. Their first product was not Cursor. It was a set of mechanical engineering tools that Truell would later describe simply as "wandering in the desert." Eventually they found their way back to what they actually knew: code editors, AI, and the gap between the two.

An editor that makes editing feel beside the point

Cursor began as a fork of Visual Studio Code - leveraging the vast ecosystem of extensions, keybindings, and developer familiarity that Microsoft had spent a decade building. This was a pragmatic choice that gave it an instant installed base and near-zero switching friction for the millions of VS Code users worldwide.

But the fork was always a starting point, not a destination. Cursor 3, launched April 2026, is built from scratch. The interface is no longer organized around files and folders. It is organized around agents - parallel AI processes that can run locally, in cloud environments, on remote SSH machines, and via integrations with Slack, GitHub, and Linear simultaneously. The IDE is still there. It is just no longer the main character.

Three years. One record after another.

The revenue chart looks like a physics violation

The customers read like a conference keynote

Cursor's enterprise pivot accelerated faster than any comparable SaaS company in the record books. When the company reached $400M ARR in late 2024, enterprise customers contributed roughly 25% of revenue. By February 2026 at $2B ARR, that share had grown to 60%. The individual developer subscriptions that launched the company are now the smaller half of the business.

When Patrick Collison of Stripe endorses your product, it is table stakes. When Jensen Huang deploys it to 40,000 engineers at the world's most important semiconductor company, it is proof of something more durable: that Cursor holds up at scale, under pressure, in production. Y Combinator's internal adoption going "from single digits to over 80%" in a short window is the kind of number that marketing departments invent and customers actually generate in this case.

$3.2B raised. Each round came faster than the last.

Total raised: ~$3.37B. Valuation at Series D: $29.3B. Reportedly in discussions for a new round at $50B+ as of early 2026.

They want codebases to drive themselves

Cursor's stated mission is to accelerate human software development through AI. The vision is considerably more radical: a future where software engineers do not write individual lines of code at all, but instead orchestrate fleets of agents that write, test, review, and deploy code autonomously. "We'll invest in the IDE until codebases are self-driving," the founders wrote on the Cursor 3 launch page - a line that reads either like a promise or a threat depending on whether you enjoy writing code.

This positions Cursor not as a code editor company but as an applied AI research organization that happens to ship a product people use every day. The in-house Composer 2 model is the first evidence that Anysphere intends to be a model company, not just a product that routes to other companies' models. By late 2025, Composer 2 generated more code than almost any other LLM in the world.

Back to that terminal window in Santa Clara

Forty thousand engineers at Nvidia open Cursor to start their day. Most of them are not thinking about the $29 billion valuation behind the interface, or the four MIT graduates who spent a year lost before finding it, or the fact that the editor they are using contains a model that now generates more code than almost any LLM ever trained. They are thinking about the bug they need to fix before the 10am standup.

And Cursor fixes it. Or at least, Cursor's BugBot found it first, left a comment on the pull request, and linked directly to the relevant line. The engineer opens Cursor, reads the comment, approves the suggested fix, and moves on. Five minutes saved. Multiplied by 40,000. Every day.

This is the actual story of Cursor: not a moonshot, not a manifesto, but an accumulation of five-minute wins that add up to the fastest revenue growth in the history of enterprise software. The IDE was never the point. The point was what happens when the cursor stops blinking and starts thinking.

With $6 billion in ARR forecast for end of 2026 and a potential new round at twice the previous valuation, Cursor's next chapter will likely determine whether AI coding tools are a feature category or a market unto themselves. The founders are betting on the latter. The Fortune 1000 is betting with them.