Sep 3, 2025 · 27m · y-combinator
Michael Truell: Building Cursor At 23, Taking On GitHub Copilot & Advice To Engineering Students · Y Combinator
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this Y Combinator AI Startup School talk, Michael Truell, CEO and Co-founder of Cursor, shares his journey from early high school AI projects to building a leading AI code editor that scaled from $1 million to over $100 million in revenue. He details critical product pivots, technical insights on in-house model training, lean startup execution, and his vision for the future of AI-driven software development.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The partners hold 16.4% of the talking time here. How this is scored →
The partners as informed peer 3.0 Guest teaching 3.0 Guest disagreement 1.4 The partners pushing back 1.1
speaking balance: gold is the partners, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 12:22 Reframing the 'Bold Move' Narrative
Truell rejects Hu's heroic depiction of making a bold corporate move, clarifying that as broke founders in a living room, they had nothing to lose.
Hardest push from the partners ▶ 11:04 Challenging GitHub Copilot's SupremacyHu pushes on why the team chose to take on GitHub Copilot when the tech industry widely assumed the coding assistant market was already settled.
Biggest teaching moment ▶ 18:35 Why Following Vocal Early Users Can Ruin ProductsTruell educates on product strategy, detailing how following the loud demands of non-coders or framework-specific developers would have compromised their horizontal vision.
The partners hold their own ▶ 21:11 Citing Internal YC Batch TransformationHu demonstrates deep ecosystem insight by citing concrete YC batch data showing Cursor's viral surge from single-digit adoption to 80% market share.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The partners as informed peer | Guest teaching | Guest disagreement | The partners pushing back | Why |
|---|---|---|---|---|---|---|
| Title Sequence: Y Combinator Presents AI Startup School | 3 | 3 | 1 | 1 | Diana Hu guides Michael Truell through his background and asks about implementing neural networks from scratch. Michael modestly explains the technical constraints of microcontrollers that forced him to write custom algorithms without standard libraries. | |
| Founding Anysphere and Initial Product Explorations | 3 | 2 | 1 | 1 | Hu inquires about Anysphere's initial ideas after graduating from MIT. Truell candidly describes their failed attempts at building CAD copilots and encrypted messaging tools, noting none of them gained traction. | |
| Pivoting to Code Completion and Challenging GitHub Copilot | 3 | 3 | 2 | 2 | Hu probes the decision to compete against incumbent giants like GitHub Copilot. Truell explains their contrarian conviction that existing players were only building incremental tools rather than preparing for full software development automation. | |
| Building and Shipping the First Version of Cursor | 3 | 3 | 2 | 1 | Hu frames the pivot as a bold leap of faith, which Truell downplays as desperate founders hacking in a living room. He walks through building their initial custom editor from scratch within three months. | |
| Key Product Insights and Transitioning to a VS Code Fork | 4 | 4 | 1 | 1 | Hu highlights the transition from scratch-built software and points out their model inspiration from OpenAI's Codex. Truell breaks down the Chromium analogy that led them to fork VS Code and discusses the unit economics of model training. | |
| Y Combinator Application Segment | 2 | 3 | 1 | 1 | Hu prompts on the uncertainty during 2023 before finding rapid product-market fit. Truell explains how following vocal user feedback gradients would have led them astray into niche or non-coder products. | |
| Hypergrowth in 2024 and Compounding Product Quality | 4 | 3 | 2 | 1 | Hu shares empirical YC batch statistics showing Cursor adoption jumping from single digits to 80%. Truell details how relentless focus on core product quality vastly outperformed dedicated growth engineering sprints. | |
| Operating Lean with a Small Core Engineering Team | 2 | 3 | 1 | 1 | Hu asks about maintaining a lean team and the long-term outlook for software engineers. Truell outlines Cursor's pragmatic 'messy middle' thesis, contrasting it with both incremental plugin approaches and pure AGI hype. |
Statements from this episode (12)
Truell: Cursor began as an AI copilot for mechanical engineers in CAD
“The first real idea that we worked on for a long time was in mechanical engineering. It was trying to build a co-pilot for mechanical engineers and trying to train models to kind of predict what you would do in a CAD system like SolidWorks or Fusion three 60 w…”
Truell: All Software Development Will Flow Through Models Within Five Years
“And I think we took a step back and realized that if we were being really consistent with our beliefs, you know, there was going to be an opportunity for all of coding to change in the next five years and for all of software development to flow through models.”
Truell: Cursor shipped to GA roughly three months after starting code
“It took us Roughly, I think, three months from first line of code to open it up and GA it.”
Truell: Cursor switched to forking VS Code to focus on AI features
“We thought that, you know, VS Code had been developed over the course of 12 years, was one of the earliest TypeScript projects had lots of people on it. Thought, oh yeah, of course, you can kind of spin something up that's just equivalent for the world in, in …”
Truell: OpenAI's Original Codex Model Cost Around $100,000 to Train
“By our calculations, Codex, which was the first, this was the first autocomplete model behind GitHub Copilot, it didn't really cost that much money to train, even though, even back then, at kind of the beginning and middle of 2022, people were talking about ho…”
Truell: Standard customer interview frameworks fail for AI coding tools
“I think that there are probably some markets where you're really well served by going and immediately talking to a bunch of people, listing down their problems really rigorously, or, you know, really kind of systematically and exhaustively thinking through eac…”
Truell: Cursor resisted vocal early user demand to target non-programmers
“Of our early user base, if you just kind of followed the gradient of exactly what they wanted, you would get pulled in slightly different directions than we ended up in. You know, we had a really loud segment of users that didn't know how to code at all, and w…”
Truell confirms Cursor reached roughly $1M in revenue around 2023
“You went from zero to one million around twenty-twenty-three, right? And it was, ah, it took a lot to get there, right? Yeah, it was a bit more than that, but sort of roughly that.”
Hu: Cursor usage in YC batches jumped from single digits to 80%
“I remember in 2023, I think it was maybe single digit percentage of the batch we used cursor, then 2024, it was like 80%.”
Truell: Cursor ended 2023 with fewer than 10 employees
“I think we ended. 2023 at only single digits people. Like we were less than 10 still.”
Truell: Fully transforming software engineering with AI will take decades
“AI is going to be this transformative technology maybe more so than, you know, any revolution in recent technological revolution in recent centuries, but it's going to take a couple of decades, and it's going to be this industry-wide effort where there are all…”