The Cursor Origin Story: From a CAD Idea to a $60B Acquisition

Cursor is the AI code editor a million-plus developers now open every day — but it started as something else entirely, built by four college students who almost aimed it at a completely different industry. The Cursor origin story runs from a scrapped mechanical-engineering idea to a fork of Visual Studio Code, an OpenAI seed check, some of the fastest revenue growth in software history, and a $60 billion acquisition. Here's how it actually happened, and the one lesson in it that matters for anyone who depends on a tool they don't own.

Who created Cursor?

Cursor is built by Anysphere, a company founded in 2022 by four MIT classmates — Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger — who met while studying computer science and mathematics. They weren't strangers who answered a job posting; they were friends who had already collaborated on research and projects together at MIT.

The four brought unusually deep technical backgrounds. Truell became CEO; Asif, who grew up in Pakistan and represented his country at the International Mathematical Olympiad, became CPO; Lunnemark won medals at the International Olympiad in Informatics; and Sanger came from medical AI and enterprise ML. Between them they'd interned at Google, Stripe, Jane Street, Two Sigma, and IBM — the kind of résumés that make a pre-seed round easy to raise.

The idea that almost was: CAD autocomplete

Here's the part most people don't know. Cursor wasn't the first plan.

The founders originally set out to build AI tools for mechanical engineering, focusing on autocomplete for CAD software. The thesis was reasonable — bring modern AI assistance to an industry that hadn't seen much of it. But the reality didn't cooperate. They quickly realized the market was stagnant, uncompetitive, and outside their direct expertise, and that data limitations and a lack of accessible training material made developing robust models challenging.

So they pivoted to the thing they knew best: writing code. By mid-2022, they moved to software development after identifying the limitations of GitHub Copilot as a plugin — a tool that had improved only incrementally over time. That reframing — the problem isn't the AI, it's the architecture — became the whole company.

The key insight: don't build a plugin, build the editor

Copilot lived inside your editor as an extension. That meant it could only see and touch what the editor's API exposed to it. The Cursor founders bet that real AI assistance needed to live underneath the editor, with full access to the codebase, the files, and the workflow.

Cursor was conceived as a platform the founders themselves would want to use — a tool that could provide meaningful AI assistance throughout the coding workflow, not just in isolated moments. They recognized that even without breakthroughs, AI capabilities would keep improving through scaling models and training data. In other words: bet on the models getting better, and build the container that would make the most of them.

Why they forked VS Code

Building a full IDE from scratch would have taken years and asked developers to abandon everything they knew. So they didn't. Instead of building a new IDE from scratch, they forked Visual Studio Code — which let them offer a familiar interface while embedding AI deeply into the development process.

The payoff of the fork is subtle but huge. Cursor is a fork of Visual Studio Code, so it inherits VS Code's Electron structure, the Monaco editor, and the entire extension ecosystem — meaning existing users could migrate settings and extensions and feel at home on day one, while the founders spent their engineering effort on the AI-native layer on top. That layer includes a retrieval system that indexes your codebase so the AI can navigate large repos instead of just the file in front of you.

OpenAI's seed check and the March 2023 launch

With a $400K pre-seed raised in April 2022, the team started building. They raised the pre-seed round to start development and later graduated from OpenAI's accelerator program in 2023, launching Cursor in March 2023.

Then came the signal that mattered. In their first year, they raised $8 million in seed funding, led by OpenAI's Startup Fund — with angel investors including former GitHub CEO Nat Friedman and Dropbox co-founder Arash Ferdowsi. OpenAI backing an AI-first developer tool was more than money; it was validation of the thesis that the future of coding would be AI-native, not AI-as-an-add-on.

From launch to $1B ARR: the growth almost nobody predicted

What happened next is why Cursor keeps getting called one of the fastest-growing startups ever. A rough timeline:

• End of 2023 — an $8M seed round and roughly 30K daily active users.
• November 2024 — Cursor acquired Supermaven, an AI code-completion startup, to strengthen its autocomplete.
• January 2025 — Cursor reached $300 million in annual recurring revenue just two years after launch, having crossed $100M ARR reportedly without marketing spend.
• April 2025 — over 1 million daily active users, 50K businesses, and roughly 9,900% year-over-year ARR growth (a figure the company reported for its December 2025 snapshot).
• Late 2025 — a US$29.3 billion valuation and more than US$1 billion in annual recurring revenue by late 2025.

Along the way, Cursor did what fast-growing tools do: it kept shipping. In October 2025 it launched Cursor 2.0 with its first proprietary AI model, Composer, described as roughly 4x faster than similarly intelligent models and built for low-latency, multi-step agentic coding. That in-house model was a strategic move to reduce dependence on OpenAI and Anthropic for the core experience.

The plot twist: OpenAI tried to buy it, then SpaceX did

The origin story has a corporate-drama third act. OpenAI reportedly tried to acquire Anysphere before its valuation skyrocketed; when that failed, OpenAI pivoted to trying to acquire Cursor's competitor, Windsurf, instead. That pursuit collapsed too — Google struck a deal for Windsurf's leadership, and Cognition (maker of Devin) absorbed what remained.

Cursor stayed independent — until it didn't. In 2026, SpaceX moved in. According to public reporting, SpaceX agreed to acquire Cursor (Anysphere) in an all-stock transaction valuing the company at $60 billion and placing it under its SpaceXAI unit. The acquisition closed in August 2026, making Cursor a wholly owned subsidiary of SpaceX, integrated within SpaceXAI.

So the tool that four students built to escape a stagnant CAD market ended up, four years later, inside a rocket company's AI division.

A note on sources: much of the funding and acquisition timeline above traces to reporting from TechCrunch, Bloomberg, and company filings. Growth figures are self-reported by Cursor unless otherwise noted, so treat exact percentages as directional rather than audited.

What the origin story actually teaches

Strip away the valuations and the drama, and Cursor's arc is a clinic in a few durable lessons:

• Architecture beats features. The founders didn't beat Copilot with a better autocomplete; they beat it by rejecting the plugin model entirely and owning the editor.
• Pivots are normal. The first idea (CAD) was wrong. The willingness to abandon it fast is what made the second idea possible.
• Bet on the trend line. They assumed models would keep improving and built to capture that — instead of waiting for a breakthrough.

But there's one more lesson, and it's the uncomfortable one. The same story that makes Cursor inspiring also makes it a cautionary tale for the people who use it. A tool you rely on every day can change owners, pricing, and model access on someone else's timeline. Cursor's users didn't get a vote when it was acquired — they got an announcement.

Own the workspace, not just the seat

The through-line of Cursor's origin story is ownership: the founders won by owning the editor instead of renting space inside someone else's. The same logic applies one level up, to you. If your most-used AI tool is one acquisition away from a pricing or policy change, the durable move is to own your workflow instead of renting a seat.

That's the idea behind Eigent, an open-source, multi-agent "Cowork" desktop app — an AI workforce that runs locally on your machine. You bring your own model keys (Claude, GPT, Gemini, or open-weight), your data stays local by default, and there's no vendor to acquire it out from under you. If you're rethinking a proprietary editor after the SpaceX deal, start with our free, open-source Cursor alternative breakdown, wire up a workflow you run daily like reviewing GitHub PRs, then download Eigent and make it yours.