With 12 engineers and a $19 billion valuation, AI Dark Horse's dream story

The development of generative AI in 2024 is described as crazy, with the big model war spreading to various tracks and the vertical application boom. Especially in the field of AI programming, algorithmic progress soared, multiple emerging unicorns were born, investors poured in intensively, and billion-dollar financing was announced from the beginning of the year to the end of the year.

But if nominated for AI Programming Dark Horse of the Year.CursorDefinitely a big hit.

In just the last six months, it's taken the programming world by storm by making writing code easy, converting tons of developers into hardcore fans. Engineers and researchers from international tech giants and top research labs, as well as "top-tier" generative AI companies such as OpenAI, Midjourney, and Perplexity, are all paying to use the tool.

How hot is Cursor? Basically, it's been praised all over the internet, and everyone who has used it says it really smells good.

Cloudflare VP Ricky Robinett posted a video showing his 8-year-old daughter building a chatbot with Cursor AI in just 45 minutes. Cursor's "kitten light" app, which took only 1 hour to develop, has quickly rushed to the Top 20 of Apple's iOS app charts, and Andrej Karpathy, co-founder of OpenAI and former director of AI at Tesla, has also waved the flag for this rising star of AI programming, boasting that Cursor's experience has already been a success. He also cheered for this rising AI programming star, quipping that Cursor's experience has crushed GitHub Copilot.

The best part is that it attracts a large number of subscribers who willingly pay for it with a pricing much higher than GitHub Copilot's $20 per month, which shows how much it hits the pain and pleasure points of programmers.

▲Cursor official website (Source: Cursor)

　　So how good is Cursor?

Simply put, it changes the way programming is learned, making line-by-line code knocking a thing of the past and turning programming into a thing that can be done with natural language communication. You come up with the ideas, and AI is responsible for helping you realize them. Backed by Claude 3.5 Sonnet, GPT-4o and other third-party big models, Cursor AI can automatically generate code, check and fix errors, and also instantly answer questions and solve puzzles when you encounter programming problems.

Not only is it beginner-friendly, but it is also a time-saving asset for professional software developers to significantly speed up their workflow.

Its investors are gathered a group of celebrities, there is Google AI leader Jeff Dean, there is OpenAI co-founder John Schulman, there is the father of derby AI Noam Brown, and former GitHub CEO Nat Friedman, Stripe co-founder Patrick Collison, Dropbox co-founder Arash Ferdowsi, among others. The latest valuation reached $2.6 billion (about Rs 19 billion).

And the startup behind the product that's racking up explosive word-of-mouth has only 12 engineers and researchers as of August 2024, less than two years old.

How did they do it?

Let's take a look back at the birth of one of the breakout AI native products of the year.

 

First, from OpenAI "graduation", 4 months valuation rose 550%, the rise of an AI code editor

In 2022, MIT students Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger started a company called Anysphere.

▲ The four co-founders of Anysphere pose for a photo (Source: Forbes)

　　The four young men had an ambitious ambition: "to create a magical tool that will one day write all the software in the world."

Their ultimate goal is to solve the hassle of building and changing software.

In its second year of business, Anysphere graduated from the OpenAI Accelerator Program and closed an $11 million seed round led by the OpenAI Startup Fund.

By the second half of 2024, the process of capital transactions suddenly accelerated.Anysphere first announced the completion of the $400 million A round of financing in August, and then announced the acquisition of Supermaven, the company behind the AI code model Babble, in November, and then in December was exposed by the foreign media to the completion of the $100 million in new financing, and the valuation of the company jumped from the $400 million in August to 2.6 billion dollars (about RMB 19 billion yuan), a full 5,50%.

One of the possible reasons for the valuation surge is tied to its fast-growing sales.

The strong demand for AI programming has translated into a strong willingness to pay. According to TechCrunch, Anysphere's annualized recurring revenue (ARR) grew more than 10x from April to October to $48 million. According to another prediction by research firm Sacra at the time, Anysphere's ARR is expected to exceed $65 million in November, a 64-fold year-on-year increase, and monthly revenues have reached $4 million (roughly Rs. 29.2 million).

▲Anysphere ARR prediction (Source: Sacra)

　　Anthropic's software development and code generation business saw a 10x increase in annualized revenue in three months, according to The Information. This comes on the heels of Microsoft's July earnings report revealing GitHub Copilot's annualized recurring revenue (ARR) of $300 million, which accounted for 40% of GitHub's overall revenue growth.Cursor's growth rate looks phenomenal by comparison.

AI Programming ToolsThe subscription model is commonly used, and Cursor is no exception.

Its pricing isn't exactly approachable, though, with a two-week free trial period, after which users have the option of purchasing the Pro version for $20 (roughly Rs. 146) per month, or a business subscription designed for large teams and organizations for $40 (roughly Rs. 292) per month.

Although Cursor charges more than GitHub Copilot's $10 (roughly Rs. 73) per month, it still attracts more than 30,000 customers around the world, including well-known companies such as OpenAI, Midjourney, Perplexity, Replicate, Shopify, and Instacart.

Feeling threatened by the competition, Microsoft quickly launched a free version of GitHub Copilot. Free users are entitled to 2,000 code completions per month. Microsoft CEO Satya Nadella also personally tweeted in support of Copilot.

Just as Cursor is taking off in 2024, the combination of AI and programming has become a hotspot for venture capital. Apart from GitHub Copilot, the mainstream preferred choice of developers, numerous other startups in the market such as Augment, Codeium, Poolside, Magic AI, Bolt.new, etc. are emerging from the fierce competition and pocketing big funding.

And how does Cursor stand out?

 

Second, challenge Microsoft to change the way software is built with AI

The core functions of AI programming can be divided into two main categories: automatic code generation and code completion/suggestion.

Cursor is an offshoot of VS Code. As a standalone native IDE, Cursor has greater control over the editor's user interface, allowing for stronger AI integration.

Unlike GitHub Copilot which focuses on providing code suggestions and completions based on context, Cursor is designed to be able to provide intelligent code hints based on the entire code base. As a result, Cursor can anticipate developer needs in real-time, and with a simple press of the Tab key, the AI editor predicts and autocompletes code, recommending the next possible code edits based on context.

▲ Analyzing the project code base (image source: Cursor)

　　Cursor's vision is to build a code editor that is more helpful and more fun than ever in the next few years.

VSCode used to be the IDE of choice for most developers, but the new age of AI calls for a new tool that interacts more naturally and integrates seamlessly into the workflow, and while GitHub Copilot serves as an enhancement to existing IDEs, Cursor builds a more specialized IDE that deeply integrates with AI.Standing on the shoulders of cutting-edge third-party big models, the Cursor team focused their product polishing on optimizing the development user experience and the way models interact, and how to make models give better answers.

In 2024, more and more developers are turning to Cursor, an AI programming tool that focuses on the seamless integration of AI into the development process than traditional IDEs. Cursor's core strengths include not only code completion and conversational programming, but also the introduction of the "Composer" feature, which allows users to make cross-file edits across the entire project by making requests in natural language. With the acquisition of Supermaven, Cursor further strengthens its technological capabilities in handling complex code and improving context-awareness.

1. Agent mode and Yolo mode: enhancing automated task execution

On November 24th, Cursor released version 0.43, which introduces "Agent Mode". This mode supports interacting with the Agent through Composer, selecting contexts and executing terminal commands. agent mode not only supports semantic search, but also helps developers to automate tasks such as code completion, bug fixing, and multi-file operations. Through semantic search, Agent mode can intelligently recommend solutions to further improve work efficiency.

On December 17th, Cursor released its latest version 0.44 and introduced Yolo Mode on top of Agent Mode.The goal of Yolo Mode is to allow agents to execute commands more intelligently and support parallel processing of multiple tasks. For example, a developer can instruct an agent to make code changes in multiple files at the same time, or perform complex terminal operations without human intervention. With the GPT-4 model, the Yolo pattern enables agents to understand commands more accurately and execute them automatically.

2、Intelligent code completion and automatic generation

One of the biggest highlights of Cursor is its deep learning capability based on advanced AI models such as GPT-4, Claude 3.5 Sonnet, etc., which is capable of predicting developers' requirements in real time and generating corresponding code. This enables Cursor to not only automatically patch, fix bugs and refactor code, but also generate new code snippets, dramatically improving development efficiency.

Compared with traditional VSCode or GitHub Copilot, Cursor is much better at contextual understanding. It is able to intelligently predict the developer's programming intent, not only relying on code snippet matching, but also focusing on the logic of the overall project, truly realizing code intelligence from a global perspective.

Cursor's Tab feature is a powerful and intuitive AI-assisted programming tool that enhances auto-completion. Not only does it auto-complete code based on the cursor position, it also supports multi-line completion and code rewriting. Cursor Prediction is another advanced feature of Cursor, which can predict the developer's next action based on the context, greatly improving development efficiency.

▲Tab key auto-completes the code (Source: Cursor)

　　3. Context-awareness and project-wide support

Cursor breaks through the limitations of traditional IDEs that are limited to the current file and supports analysis of the entire project. It understands project structure, coding style and provides code suggestions based on a global view. @Codebase is a powerful feature of Cursor, which analyzes the entire project through vectorization techniques to provide developers with comprehensive contextual support, especially for large projects or scenarios that require cross-file analysis.

With natural language queries, developers can easily issue commands such as "change all functions in userService" or "add unit tests for this module" and Cursor will automatically generate the appropriate code snippet. Compared to GitHub Copilot, Cursor has the advantage of global project understanding and cross-file coordination.

▲Edited in natural language (Source: Cursor)

　　4. Composer: intelligent coding assistant

"Composer is Cursor's intelligent coding assistant designed to help developers easily write and modify code. Users simply interact with Composer through natural language commands, such as "add login" or "optimize this function", and Composer automatically generates and optimizes the code. This greatly simplifies the programming process and allows developers to focus on higher-level design without having to switch contexts too often.

5、Cursor Chat: Interactive Code Searching

Cursor Chat provides an interactive code search feature that allows developers to interact with the code base through dialog. Users can enter simple natural language questions and Cursor Chat will provide relevant code snippets or explanations based on the context of the project.Cursor Chat is unique in that it combines contextual references (@) with the "Apply" functionality for code modifications, which allows developers to apply code suggestions provided by the AI directly to the project. This allows developers to directly apply the code suggestions provided by AI to their projects.

While similar conversational programming tools already exist, Cursor Chat simplifies the code modification process with its "Apply" feature, providing a more efficient solution than traditional code editors.

6. Privacy and security

For developers, the privacy and security of their code is of paramount importance, and Cursor provides strong privacy safeguards in this regard, ensuring that code is not stored on servers through privacy mode and enhancing data protection through SOC 2 certification. All code processing is done locally to ensure that developers' data security and privacy are not violated.

In short, Cursor drives the evolution of IDEs through its deep integration of AI features, intelligent code completion and generation, semantic search, and intelligent task execution in Agent and Yolo modes. Especially in terms of full project support, context-awareness and privacy protection, Cursor provides a more efficient and intelligent solution than traditional IDEs.

 

III. Copilot, Cursor, and Bolt.new: three development paths for AI programming tools

Globally, AI coding tools such as GitHub Copilot, Cursor, Augment, Codeium, etc. have been widely used in the market, and domestic vendors such as Aliyun, ByteDance, Huawei, etc. have launched their own AI programming products, such as Alibaba's Tongyi Spirit Code and Baidu's Wenshen Quick Code.

Using data provided by research firm Sacra, Twitter user auchenberg created a chart comparing the annualized recurring revenue (ARR) of the major AI programming products currently on the market. You can see that Cursor's ARR jumps a ton from 2023 to 2024, but the overall volume still doesn't compare to GitHub Copilot.

▲ ARR of mainstream AI programming products on the market (Source: Twitter user auchenberg, Sacra)

　　Among them is GitHub Copilot's ARR, which grew from $115 million in 2023 to nearly $400 million in 2024, a growth rate of 2,811 TP4T. Bolt.new's ARR of $20 million by November 2024 and the platform's average daily ARR growth of $500,000 are also growing nicely.

GitHub Copilot, Cursor and Bolt.new represent different development directions for AI programming tools. GitHub Copilot focuses on mass popularization and multi-platform integration, Cursor is an intelligent upgrade based on existing IDEs, and Bolt.new explores a new model for web development.

▲ Different types of code editors (Source: Cursor)

　　GitHub CopilotAs a product of Microsoft, it directly integrates two platforms, Visual Studio Code and GitHub, providing powerful code-completion features and supporting more than dozens of programming languages. Trained by deep learning models, Copilot is able to understand context and generate reasonable code snippets, greatly improving development efficiency.

CursorOn the other hand, it focuses more on integrating AI functions on the basis of traditional IDEs (such as VSCode), and helps developers improve coding efficiency by providing intelligent code completion, automated refactoring and other functions. Its success lies not only in its efficient auxiliary functions, but also in the good user experience and ease of operation. Users can get real-time advice and support while writing code, reducing the learning curve for junior programmers while providing an efficient tool for senior developers.

Unlike the previous two local IDEs, theBolt.newThe pioneering realization of browser-based low-code development provides a more intuitive, low-code development experience, relying on WebContainers technology to run a miniature operating system in the browser and integrating the Claude 3.5 Sonnet model. This allows users to code directly in the browser without configuring a virtual machine or logging into an account, simplifying the development process and making it especially suitable for beginners and rapid development needs.

Cursor improves development efficiency through a global project view and intelligent task execution, and Bolt.new lowers the technical barrier by simplifying the development process. The success of these products reflects the strong demand for intelligent tools from developers, who are no longer satisfied with the single functionality of traditional IDEs, but are more concerned with tools that are tightly integrated with business needs.

 

Conclusion: AI programming development on the fast track

AI programming tools bring advantages along with challenges. Code quality and security have become important issues.AI-generated code may contain errors or inconsistencies and still need to be reviewed manually. In addition, data privacy and intellectual property issues need to be bound by compliance. In the future, how to balance technological innovation with the guarantee of code quality and security will be the key to the development of the industry.

What's more interesting is that Cursor publishes a blog post each year describing a list of problems it hopes to solve. 2023 lists problems such as better context, editing Copilot, proactive error-checking, and larger editing scope, and 2024 lists include next-action prediction, perfect editing, best context, error detection, and debugging.

Going into 2025, the AI programming track is expected to grow even more by leaps and bounds. After all, the Cursor team has already set the flag in an interview, "The Cursor of a year from now is going to make the Cursor of today look obsolete."

Note: This article is from "Smart Stuff", author | Wang Yue