Cursor: An AI Code Editor Enriches the World

An 8-year-old girl sits at her computer and starts programming. As if it were the easiest thing in the world. She is the daughter of Cloudflare Vice President Ricky Robinett and is using a chatbot and does coding “simply [by] typing text-based prompts into an AI coding tool called Cursor” [1]

This Cursor-app was released back in June, but has only recently gained widespread acceptance in the tech scene. It is not often that an app gains general acceptance and broad approval from the AI and developer scene within a short space of time. Numerous different tools are sometimes circulating the web [2] - but this time it seems to be different. There is widespread unanimous support for Cursor and it is described as the tool par excellence for programmers. “Engineers at leading AI startups like OpenAI, Midjourney, Perplexity and Scale AI, as well as tech companies like Shopify and Instacart, are among the 30,000 customers who pay $20 or $40 per month to use Cursor's AI tools to write and edit entire chunks of code” [3]

Searches for Cursor have exploded since the end of July and beginning of August.

The success of Cursor has even attracted Forbes and prompted the editors to write a major article about the app - of course only in the highest terms and almost a eulogy for the app's efficiency.

When ChatGPT saw the light of day in 2022, it had an unprecedented impact on the way we generate knowledge. Today, we can say that Cursor has done the same with programming. Since the advent of LLMs, former sites like Stackoverflow have fallen behind, especially after Microsoft released the GitHub CoPilot. However, the more Cursor pushes itself into people's focus, the more so. I would therefore like to investigate the question:

What Is Cursor and What Makes It Successful?

In 2022, the company Deepgram was founded by four friends who met at MIT and within a short space of time received around $60m in seed capital for their start-up from well-known names such as VC Andreesen Horrowitz, OpenAI Startup Fund and others. The company has now grown to a valuation of $400m. The friends started out with the aim of developing an app that would benefit beginners and professionals alike, make everyday coding easier and multiply productivity. But the result is even more than that.

Programming Without Prior Knowledge

First of all, you can call Cursor an AI code editor. It is Visual Studio on AI steroids and combines numerous useful features of a GitHub CoPilot in one app and even improves them in key areas by incorporating additional efficiency features. In this way, Cursor shows how AI can already pragmatically change and improve everyday working life. Although Cursor is not the big breakthrough of an AGI, ASI or Singularity (unfortunately I have to emphasize this, because some people only listen to trigger words like AGI/ASI and Singularity, but unfortunately pay little attention to the daily iterative developments), Cursor still gives an essential insight into what AI can already do today to give our everyday productivity a big boost thanks to this technology and, like a telescope, a glimpse into the seemingly distant but already imminent future. What's more, Cursor is leading the way in the current use of AI to the extent that language models are so deeply integrated into the editor that the sum of its parts is more than just editor+AI. The Language Models are so interwoven with the code editor that programming is possible without prior knowledge (Two examples of children learning to program with cursors [4.2]). But at the same time so complex that veteran programmers swear by it.

Forbes summarizes it this way:

The result is also an application that is both child's play to use and enables maximum complexity

Diverse Application Possibilities

Let's first take a look at what Cursor can do and how it is superior to other similar tools.

AI-supported code generation and completion: Cursor.ai provides intelligent code completion and generation that is tuned to the context of the code. This allows developers to work faster and spend less time on repetitive tasks. This feature is particularly useful when developing large code bases. ("Contextual Understanding: Cursor understands the specifics of your repository, allowing you to ask questions or get suggestions relevant to your project.” [7] ; ”In addition, Cursor offers intelligent features such as Smart Rewrite and Prediction, which predict and suggest the next section of code based on your current context. These features speed up the development process and significantly improve the accuracy and quality of the code. It's like having an assistant that not only understands what you're doing, but also what you should do next.” [8]

Codebase chat and in-context help: One of the most outstanding features is the “Code Chat”, which allows developers to ask questions about their code directly to the AI. The AI can access the entire codebase and provide contextual answers. This reduces the need to consult external documentation ("Speed and Efficiency: By aiding in documentation browsing, code changes, bug fixes, and even generating code from scratch, Cursor significantly speeds up the development process.” [9])

Automated debugging and linting: Cursor.ai provides automatic error detection and debugging help by highlighting potential problems in the code and suggesting solutions. In addition, the AI can automatically fix linting problems, which significantly improves code quality.

The Multi-Line-Edit feature of Cursor.ai again allows developers to edit multiple lines of code simultaneously. This is especially useful for repeated changes in multiple lines, such as inserting prefixes, deleting or replacing text in a single step. The feature saves time and reduces errors by allowing multiple lines to be edited in one step, rather than changing each line individually. It is seamlessly integrated into the editor and contributes to overall efficiency and ease of use. Cursor.ai can automate this task by performing the change synchronously in all selected files. “Especially for large projects, this is a huge benefit as the composer integrates variables, components and styles consistently across all code. Supported by powerful AI models such as GPT-4, Composer enables unprecedented efficiency and precision. This feature makes Cursor one of the most effective AI-based development tools on the market.” [10] Autocomplete and automatic error correction, context-based code generation, multi-line edit and much more are already helpful supporters individually, but because they can all be used together in an app with an understanding of the context, Cursor is - I repeat - more than the sum of its parts. Within seconds, entire web applications can be created, the code generated, optimized, refined and automatically deployed - across multiple contexts.

Some features “come in Handy”, others are outstanding (some more have been described very well on deepgram.com, I have added them to the footnote FN) [13]

In short, Cursor's uniqueness lies in the unification of all the AI benefits for coding in a single app; it brings the groundbreaking support of AI directly into the development environment and lets you stay in the flow of development. It is also designed to be easy to use even for beginners thanks to helpful welcome videos.

Who Is Cursor Aimed At?

The special thing about Cursor is that it is aimed at both beginners and professional developers alike. A contradiction that is otherwise almost impossible to resolve, because something is either simplistic or professional and therefore often overflowing with functionality. However, Cursor allows beginners to program using natural language input, no prerequisites in any programming language are necessary. In other words: Cursor is immensely user-friendly and has an intuitive user interface.

Professional users, on the other hand, can use the integrated AI (Sonnet-3.5 / GPT-4o, API) to either have entire code sections generated directly and thus save time or correct annoying (careless) errors directly in the development environment by the AI. In addition, McKay Wrigley has demonstrated in almost every video that he programs faster by using the voice function to command the LLM the code to be generated than by tedious manual prompting ([14]; McKay has also created numerous other great videos with Cursor, e.g. by creating a financial dashboard with Cursor within 5 minutes [15])

The company behind Cursor, Deepgram, itself writes that Cursor is aimed in particular at the following groups:

• “Software Engineers: Professionals looking to enhance their productivity and coding efficiency.

• Students and Educators: Those in educational settings can benefit from its intuitive interface and AI assistance.

• Project Managers: Individuals overseeing development projects can utilize Cursor to streamline workflows.” [16]

Or in other words: everyone. To students, specialists and managers. Cursor should be useful to everyone.

What Comes After Cursor?

I would like to say that Cursor is the first revolutionary AI tool that easily doubles, maybe even triples, productivity in development. I am very excited about the further development of Cursor and especially about the future implementation of new models (Opus 3.5, GPT-5/NEXT) that will give a further boost to coding.

But let's take a look at what comes after Cursor. Cursor is just the beginning and is still comparatively traditional in its operation. By this I mean that the AI interacts proactively in some cases and makes suggestions or even directly improves errors, but ultimately the majority of this tool is still based on input and constant exchange with the program. But what if we connect cursors with real agents? Agents that not only generate and enter the code, but also apply it, draw conclusions, continue to improve and work out solutions independently? Agents that only need to be told what you want and take all further steps on their own? That's what comes after Cursor. Cursor is currently creating the following:

“Eventually, Cursor wants to build tools that can automate 95% of the more tedious work that engineers do, allowing them to spend more time on the creative aspects of coding.“I think individual engineers very soon will be able to build systems that are much more complex than what powerful teams can build right now,” he said.”[17]

This means that Cursor offers a first glimpse into the future, in which we are developing a new relationship with devices; human-device interaction is the next big thing (keyword AR/VR). In the foreseeable future, we will probably only be programming in natural language - the agent will do everything else for you.

A video recently made the rounds in which Jensen Huang was asked on stage what he would recommend his friends to study at university. His answer was clear: definitely not programming, because in the future it would almost certainly be completely automated by AI and the only programming language that would remain would be plain English. [18] This is because AI will convert the spoken word directly into code - just as Cursor already does today, using Sonnet 3.5 or GPT-4o to understand the content of the request and generate the code based on the context. This is what I mean when I say that Cursor provides a first glimpse into the future. Programming will emancipate itself even further from human skill and will be fully mapped by language models using AI (especially because programming languages can be produced excellently by LLMs, insofar as the predicted token does not allow many alternatives; in contrast to human language, in which sentences can be diverse and alternative words variable, code is very one-dimensional, very predictable). Cursor is the harbinger of what is to come: generalized, automated code generation using prompting, self-learning and self-correcting applied by agents. That would then be the point at which we presumably also enter the singularity.

Conclusion:

Cursor feels like something revolutionary in its functionality, a novelty that represents a new benchmark in efficient AI assistants, it is a level of its own. The speed and efficiency increases many times over. “Before Cursor, it would've taken me at least 5 minutes of tracing code to get to this answer. With Cursor, it took less than 30 seconds.” [19] . Cursor is the first real glimpse into the future of coding. A future where coding is programmed in natural English. It is the current state of AI utilization to make work more efficient and productive; 2x, 3x faster work is already possible today. However, with better LLMs, a larger context window and other refinements, cursors will probably seem like a relic from the past in a year's time at the latest. The future remains exciting. But until then, we'll be using Cursor.