Episode Summary
Executive Summary: Michael Truell traces Cursor’s origin from early AI/robotics experiments to a conviction that software development will be transformed by models. After several failed side projects, the team pivoted in late 2022 to AI coding, shipped a crude editor in months, then iterated rapidly on product quality and proprietary models. Growth compounded through 2024 as the product became indispensable for builders.
Main Topics: Founder origin story and early programming exposure (Priority: 5/5): Truell describes getting into programming in middle school via mobile games, Objective-C, and inspiration from Y Combinator essays and founders, establishing the long arc behind Cursor. Early AI and robotics experiments (Priority: 5/5): Before Cursor, he and collaborators built projects like a spoofing app, robotic dog ideas, and small robots using reinforcement learning, which taught practical ML lessons and sparked long-term interest in AI. Pivot from failed ideas to AI coding (Priority: 5/5): The team tried mechanical-engineering/CAD copilot ideas and encrypted messaging before realizing their strongest conviction was AI for coding, where they believed the market was under-aimed and ripe for automation. Building Cursor’s first product (Priority: 4/5): Cursor began as a custom editor with AI features, remote SSH, and early autocomplete; the team later switched to VS Code as a base and focused on AI-first workflows rather than recreating an entire editor stack. Model strategy and product iteration (Priority: 5/5): They initially avoided training their own models, then later used product data and in-house models to improve next-edit prediction and other core features, making the product more capable as usage scaled. Growth, adoption, and company evolution (Priority: 4/5): Cursor’s usage grew from small early numbers to major traction in 2024, driven by product improvements and word of mouth rather than heavy growth engineering; the company remained very small through 2023. View on the future of coding and advice to builders (Priority: 5/5): Truell argues coding will be transformed over the next five years, but software development will remain human-in-the-loop for a long middle period; he advises young builders to work on what genuinely interests them with people they respect.
Key Arguments: Cursor’s founders believed the entire coding stack, not just small UI improvements, would be reshaped by models within five years. Early AI and robotics side projects were valuable because they forced the team to learn ML fundamentals, data efficiency, and model training through real constraints. Their first serious bets failed because they were in markets they did not deeply care about and where the science/product fit was too early or too narrow. Cursor succeeded by focusing on being the best way to code with AI for professional engineers, rather than a niche tool or a horizontal consumer product. A code editor alone was not enough; the team learned that the real leverage was in AI workflows, prediction, and action-taking inside the codebase. Building proprietary models and using product data became a key advantage as the product matured and scaled. Growth came mainly from product quality improving continuously, which immediately showed up in usage and adoption. The future of coding will be a long, messy transition where AI behaves increasingly like a colleague and advanced compiler, but code review and logic understanding will still matter. Programming and math remain valuable general education even as AI changes the tooling landscape.
Data Points: Age: 24 - Truell is described as having built Cursor at a very young age. Time from first line of code to public launch: ~3 months - Cursor’s initial editor was built and opened to users quickly. Time to daily-driver usable product: ~4 weeks - They reached an internal daily-driver version of the editor about four weeks after starting. Time to first beta testers: ~8 weeks - Four weeks to daily driver plus another four weeks before beta access. 2022 GitHub Copilot revenue: ~$100M - Used to illustrate how competitive the coding-assistant space already was. Company size at end of 2023: <10 people - Cursor stayed very small while product-market fit was still forming. Growth from 2023 to 2024: ~1M to 100M - The speaker cites an explosive year-over-year scale-up, described as roughly 1 to 100 million. YC adoption in 2023 batch: single-digit percentage - Very few YC founders used Cursor in 2023. YC adoption in 2024 batch: ~80% - Cursor usage among YC companies spread widely by 2024. Codex training cost: ~$100K - Referenced as an example that early autocomplete models were cheaper to train than many assumed.
Pivotal Quotes: "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" — Michael Truell: Explaining the conviction that pushed the team toward Cursor. "all of coding, as we know it today, gets automated and building software ends up looking very, very different" — Michael Truell: Describing the scale of change the team believed was coming. "the future of coding will be a long, messy middle where you will be working with the AI, more and more it will become like a colleague" — Michael Truell: His view of near-term software development and human-AI collaboration.
Implications: The episode suggests AI coding tools are shifting from autocomplete to deeper software creation partners. Builders should expect rapid product iteration, model integration, and human oversight to remain important during the transition.
About Y Combinator Startup Podcast
We help founders make something people want. The Y Combinator Podcast is where builders talk about building. From the earliest days of an idea to scaling a company that changes the world, YC partners and founders share real stories, lessons, and tactics from the frontlines.