Michael Truell and Cursor: Leading the Fastest-Growing SaaS Revolution in Coding

n a recent episode of How to Build the Future, Michael Truell, co-founder and CEO of Anysphere, the company behind the AI-powered coding platform Cursor, shared his vision for revolutionizing software development.

With Cursor achieving a staggering $9 billion valuation and $100 million in annual recurring revenue (ARR) just 20 months after launch, Truell’s insights offer a glimpse into the transformative potential of AI in coding. Below is a distilled version of the conversation, highlighting his key quotes and ideas.

A New Paradigm for Software Development

Truell and his co-founders, all seasoned programmers, founded Anysphere with a bold mission: to redefine how software is built. Instead of laboriously editing millions of lines of code in esoteric programming languages, Truell envisions a future where developers can simply describe what they want, and the software materializes.

“The goal with the company is to replace coding with something that's much better... We think that over the next 5 to 10 years it will be possible to invent a new way to build software that's higher level and more productive.”

This vision drives Cursor’s development. While AI already assists in writing 40-50% of code within Cursor, Truell believes the industry is far from fully realizing this potential.

For professional developers working on large, complex codebases, “vibe coding”—where developers describe high-level ideas without engaging deeply with the code—remains impractical due to the intricate, long-term nature of such projects.

“The vibe coding style of things is definitely not something that we recommend if you're going to have the code stay around for a really long time.”

Instead, Cursor focuses on empowering professional programmers by enhancing productivity while maintaining control over the code. Truell sees AI evolving from a productivity tool to something that fundamentally changes the artifact of software itself.

The Role of AI: Helper or Compiler?

Truell offers two lenses for understanding large language models (LLMs) in coding. One view is to treat them as human-like assistants, taking on delegated tasks or providing real-time suggestions. The other is to see them as advanced compilers or interpreters, translating high-level human intent into executable code.

“One way in which you can view LLMs is [to] interface with them like a human, like a helper. Another way... is they're kind of an advanced compiler or interpreter technology.”

This dual perspective shapes Cursor’s product design. Truell emphasizes the importance of giving developers precise control over details, such as moving a UI element a few pixels or tweaking specific logic.

However, he also envisions a future where the logic of software is expressed at a higher level of abstraction, reducing the need to micromanage code.

“If we were to get to a place where you don’t have to pay attention to the code as much... that written down version of the logic of the software is going to have to get higher level.”

Overcoming Bottlenecks to Superhuman AI

Truell identifies several challenges to achieving superhuman coding agents. One major bottleneck is the context window size of LLMs, particularly for large codebases that could span tens of millions of tokens.

Efficiently processing such vast contexts, both technically and cost-effectively, remains a hurdle.

“There are a bunch of bottlenecks to agents being human level... If you have 10 million lines of code, that's maybe 100 million tokens... Both having a model that can actually ingest that [and] one that actually pays attention effectively to that context window is tricky.”

Beyond technical limitations, Truell highlights the need for continual learning and long-term task persistence. Current models struggle with tasks spanning hours or requiring integration with external tools like logs or debugging systems.

These “known and unknown devils” must be addressed to make AI truly superhuman in coding.

The Irreplaceable Role of Taste

As AI automates more of the coding process, Truell believes one human quality will remain irreplaceable: taste. This encompasses not just aesthetic choices in visual design but also decisions about how software logic should function.

“One thing that will be irreplaceable is taste... defining what do you actually want to build.”

In today’s programming, developers often act as “human compilers,” translating high-level ideas into low-level code through for loops and if statements. As AI takes over this compilation step, developers will shift toward becoming “logic designers,” focusing on defining the desired outcomes and behaviors of software.

“More and more of that human compilation step will go away... but that kind of taste for what is actually useful for what you want to build, I don’t think will ever go away.”

From CAD to Code: The Cursor Origin Story

Anysphere’s journey began with an ambitious but misguided attempt to build an AI co-pilot for mechanical engineering, specifically for computer-aided design (CAD). Despite extensive user interviews and efforts to train 3D autocomplete models, the team realized they lacked the passion and domain expertise needed to succeed.

“We really weren’t as excited about mechanical engineering as we were about coding... The science back then wasn’t yet ready for 3D.”

This pivot to coding was driven by the team’s personal love for programming and the realization that AI’s scaling laws—where more data and compute reliably improve model performance—made coding a more promising domain. Inspired by tools like GitHub Copilot and early Codex papers, they saw an opportunity to redefine coding itself.

“Building a company is hard, and so you may as well work on the thing that you’re really excited about... We set off to work on the future of coding.”

Building a Durable Moat

In a competitive landscape, Truell likens the AI coding market to the search engine wars of the late 1990s or the consumer electronics boom of the 2000s. The key to a durable advantage lies in distribution and rapid iteration.

“If you keep pushing the frontier faster than other people, you can get really big gains accruing to you.”

By serving a large user base, Cursor gathers critical data on where the product excels or fails, feeding this into R&D to improve both the product and underlying models. This feedback loop mirrors how search engines refined results based on user interactions.

Maintaining the Hacker Mindset

As Cursor scales, maintaining its hacker energy is a priority. Truell emphasizes hiring passionate, multidisciplinary talent who share the company’s vision. The final stage of Cursor’s hiring process—a two-day on-site project—ensures candidates are not just skilled but genuinely excited about the problem space.

“The last step of our hiring process is a two-day on-site where you come and you just work on a project with us... That gets at energy and excitement and passion for the problem space.”

To foster innovation, Cursor encourages bottoms-up experimentation, allowing teams to explore bold ideas independently. This balance of structured coordination and creative freedom keeps the company agile.

The Road Ahead

Truell’s excitement for the future is palpable. He sees 2025 as the dawn of an “age of intelligence,” where AI will unlock unprecedented opportunities for software development. From enabling niche software for industries like biotech to accelerating the creation of new AI models, the potential is vast.

“Things that you want to happen on computers will then just kind of be able to happen.”

For Truell, the journey is just beginning. Cursor’s mission to replace coding with a higher-level, intent-driven process is a bold bet on a future where developers wield AI to bring their visions to life with unparalleled speed and precision. As he puts it, it’s about following the line—anticipating where AI’s capabilities will take us and building toward that horizon.