Michael Truell

CEO & Co-Founder — Cursor

Profile

Michael Truell is the co-founder and CEO of Anysphere, the company behind Cursor — the AI-first code editor that has eaten a large chunk of the developer tooling market in an impossibly short time. He started the company in 2022 with three MIT classmates (Sualeh Asif, Aman Sanger, and Arvid Lunnemark) after turning down the standard CS-graduate offers. The four of them bet early that transformers were about to rewrite how code gets written, and they wanted to be the editor that rewrote itself around that bet.

The origin story is worth knowing because it’s instructive: the team initially spent nearly a year building AI tools for mechanical engineers, failed at it (none of them were mechanical engineers), and pivoted. They forked VS Code, kept the familiar surface, and rebuilt the internals around AI — tab completion, multi-file edits, a chat sidebar with real codebase context, and an agent that can actually execute. Cursor launched in March 2023 and became the fastest B2B software company to reach $1B in ARR. In November 2025 they raised $2.3B at a $29.3B valuation. Truell was 25.

What makes him interesting isn’t the fundraising. It’s that he and his co-founders shipped a product that competes directly with GitHub Copilot and Claude Code — tools made by the companies that supply their underlying models — and won anyway. The moat is taste: they thought carefully about what an editor should feel like when an AI is doing half the typing, and they iterated faster than the incumbents could.

For developers learning AI, Truell is worth paying attention to because he’s one of the clearest thinkers on where coding is actually headed. His public position is that the editor is a transition artifact — eventually you describe what you want and agents handle the rest. He’s also been unusually honest about where this breaks today: he’s warned publicly that pure “vibe coding” without understanding the output builds shaky foundations. That tension — between building the thing that automates coding and insisting humans still need to understand what gets built — is the most useful frame he offers.