The episode features Michael Troll, co-founder and CEO of Cursor, in conversation with general partner Martin Cassada at a conference. Cursor is one of the fastest-growing companies in the AI coding space, having achieved unprecedented scale with a small team.

The discussion explores Cursor's origin story, including a false start as a 3D CAD company before pivoting to AI-powered code editing. Michael shares how the team was inspired by GitHub Copilot and scaling laws in 2021-2022, leading them to pursue the "cursor for X" vision across different verticals.

Martin guides the conversation through operational challenges of hypergrowth, including infrastructure scaling that stressed major cloud providers, unconventional hiring practices that persist at 200+ employees, and aggressive M&A strategy for talent acquisition.

The talk covers strategic decisions around product focus, multi-product expansion into the "AI coding bundle," and the philosophical challenge of building software tools while software itself is being disrupted by AI.

From 3D CAD to AI Code Editor

Cursor originated from a "cursor for X" vision where the founders believed there would be companies automating different verticals of knowledge work, each building the best product, winning distribution, and eventually working on underlying models to create a flywheel.

The team initially worked on mechanical engineering and CAD systems for 6-7 months but faced severe founder-market fit problems. "There was this blind man in the elephant problem where we would hop on calls with mechies and ask them what they do during their days and we never really had like an intuitive sense for it" - Michael.

The CAD space had a much harder cold start problem than coding, with no good 3D representations or out-of-the-box models that worked well. The team spent most of their time on modeling work and data scraping, creating "PTSD" that influenced their later focus on speed.

Two key moments inspired the pivot to coding: trying GitHub Copilot as "the first existence proof" of useful AI products, and excitement about scaling laws suggesting models would improve even without new ideas (around 2021-2022).

Strategic Focus Over Feature Sprawl

While competitors immediately pursued agents, custom foundation models, and multi-IDE support, Cursor focused narrowly on forking VS Code and building a superior editor experience. The team had only four co-founders and limited seed funding compared to modern rounds.

"People just thought it was very weird to do an editor... They said you can't get people to switch their code editor. They're too tied to it. Which we knew was wrong" - Michael, noting the team had switched from command-line Vim to VS Code because of Copilot.

The team built their first IDE from scratch in "a couple weeks" as a daily driver, then launched a public beta within "a couple of months" total. Monthly investor updates served as a commitment device to maintain momentum.

The founders deliberately avoided touching model development initially despite endless debates over breakfast, lunch, and dinner about core strategic questions. They wanted to "get something out to the world" before backing into model work.

Scaling Through Infrastructure Chaos

Cursor's growth was so rapid they took down a major cloud provider, and someone showed up at their office with an iPad displaying "cursor's down" on the window during a service disruption.

The tiny team operated "a very very large Kubernetes cluster larger than many other companies" with just five total people, leading to multiple infrastructure hiccups including core DNS failures and database scaling issues.

Cursor eventually comprised "a really high double digit percent" of certain API providers' revenue, forcing providers to make capacity planning and financing decisions. "The API providers really didn't know what to make of us because it's you know these four 20somes" - Michael.

The team became expert at "hunting out all the sonnet tokens that exist in the world," spreading usage across multiple token resellers and providers with committed contracts to ensure capacity.

Database scaling progressed from standard RDS instances to sharding attempts, eventually switching to PlanetScale after discovering AWS's "limitless" service still required sharding at extreme scale.

Unconventional Two-Day Hiring Process

Every engineering and design hire spends two days in the office working on real projects with a frozen version of the codebase. "Here's a desk, here's a laptop... here's three projects you could work on... just go do it" - Michael describing the free-form process.

The process tests for "orthogonal things to the normal coding style interviews," including whether candidates can "go end to end in the codebase," their agency, and product sense since "our engine design and product are pretty tightly coupled."

Cursor has maintained this practice despite growing to over 200 people and multiple attempts to kill it internally. The process provides candidates extensive information about what showing up on day one will actually be like.

Early sales hires went through similar practical tests: "Here are inbound leads and have a quota." The very first sales rep was asked to "teach us how we should do sales" with access to real data.

Recruiting stunts included flying across the world after candidates said no, then creating a fake dinner with researchers in SF six months later to reignite the conversation, successfully converting one person who became "one of the best people on the team."

M&A as Talent Acquisition Strategy

Cursor has been unusually aggressive with M&A for a 2-year-old company, contradicting the pre-AI adage that "startups should never buy startups." The primary driver is "do anything possible to get the most talented people" - Michael.

The first major acquisition was SuperMaven, a five-person team led by Jacob, who had built TabNine (the predecessor to GitHub Copilot) and was a researcher at OpenAI working with John Schulman at Thinking Machines.

SuperMaven was working on complementary autocomplete model technology. Cursor "built a relationship stayed close over many months" before aggressively approaching them about joining forces.

Looking forward, Cursor plans to use M&A strategically to build out a suite of products and create "GM type structures within the company" for complementary offerings in the AI coding bundle.

Multi-Product Vision and Go-to-Market

Cursor aims to become "the AI coding provider" for customers by building a complete "AI coding bundle," though the main focus remains the editor as "the pane of glass that you sit in when you're an engineer going about your day."

The company has expanded beyond the editor to products like BugBot and CLI, recognizing that "the ways in which work is changing within the editor start to affect how teams work together too."

"Many founders underappreciate how tough it is to go from single product to multi-product when it comes to actually go to market" - Martin, noting the complexity of cross-sell from both PLG and sales team perspectives.

The team is "still learning" how to give new product projects proper "air cover" and execute cross-sell effectively, though they're "very excited by kind of the early results."

The Long Road to Software Automation

"Despite the headlines. Despite how much demand there is in this market and how much software has changed over the last few years, it's so far away from being automated 100%" - Michael on the distance to full automation.

Building software in professional settings with "anywhere from dozens to tens of thousands of people" remains highly inefficient. "It's really easy in an executive level to underestimate just how far away we are from the limit of automating software" - Michael.

The market "had a iPod moment and like it's going to have an iPhone moment and another iPhone moment," with several such transitions already occurring and more expected in the future.

Cursor has tried to build a company that "can continually build those things because if we don't, you know, we're kaput." The rapid evolution is both a challenge and a competitive advantage against larger incumbents like Microsoft.