In the early 1950s, Grace Hopper coined the term “compiler” and created the first version with her A-0 SystemThe compiler separated out the machine code, allowing programmers to focus on high-level logic rather than lower-level hardware details, Today, AI coding assistants are enabling a similar transformation, helping software engineers focus on higher-order work by generating code from natural language promptsEveryone from big tech to well-funded startups is competing to achieve this change, tomorrow google announced AntiGravity, their new AI coding assistant, and a day ago, AWS announced General availability of their AI coding tool, Kiro. Last week, Cursor, the standout startup in this space, raised $2.3B in its Series-D round at a valuation of $29.3B.
two lines in cursor Press release Was standing in front of me. First:
We’ve also surpassed $1B in annual revenue, counting millions of developers.
This disclosure means that Anisphere Inc. (parent company of Cursor) is the fastest company in history to reach $1B in annual recurring revenue (ARR). Yes, faster than OpenAI, and even faster than Anthropic,
Engineers are trying out every new AI coding tool. As a result, the AI-coding tools market is growing rapidly (+5x in just one year)But it is still early, as i wrote Why some AI rappers create billion-dollar businessesCompanies spend several hundred billion dollars per year on software engineering, and AI has the potential to unlock productivity gains across that spend.
Software developers represent about 30% of the workforce at the world’s five largest market cap companies, all of which are technology firms, by October 2025. Development tools that increase productivity even by modest percentages unlock billions of value.
In my view, this emerging market is segmenting based on three types of users.
is at one end handmade codingThese are engineers who either actively refuse to use LLM skepticism About quality or the insistence on complete control over every code. He argues that accepting AI suggestions creates technical debt that you can’t see until it gets into production. This segment continues to decline as the quality of AI coding models improves.
is the opposite end vibe codingThese are generally non-engineers, who use AI to create concepts and prototypes, They prompt the model in hopes of an end-to-end solution, accept the output with minimal review, and trust that it works, The user describes what they want and lets the model figure out the implementation details of how to build it,
sits in the middle Architect + AI CodingEngineer uses AI/LLM as a duo programmer Exploring system design, analyzing data models, and reviewing API descriptions. Even when the work is brand new or something that needs careful handling, a human programmer codes those pieces by hand. But for boilerplate code, package installation, generic user interface (UI) components and any type of code that is commonly found on the internet, they assign it to the modelThe engineer takes control of what is important to them and delegates what is not,
Based on user types, I think, the AI coding market gets divided into two.
-
Stay away: Non-engineers (product managers, designers, marketers, other internal staff) use these tools vibe code Early product concepts. They look to the lead engineer to spin-up concepts/prototypes of apps, websites and tools by inspiring AI to build something for them. Lovable, Vercel, Bolt, Figma Make and Replit fit hereAs of now, these users’ codes are generally not shipped to the product,
-
hands on: Professional software engineers use these tools in their existing workflow to ship production code. They use AI as an assistant to write boilerplate code, refactor existing services, wire up new features or UI screens, and triage bugs in the codebase. Cursor, Cloud Code, OpenAI Codex, GitHub Copilot, Kline, AWS Kiro come into play here. These products remain where the work is doneAnd integrate into the engineer’s workflow. This is, at least so far, the larger market segment.
To see an evaluation of all the major AI coding tools currently on the market, check out Breakdown By Peter Yang, who runs the newsletter behind the craft,
This brings me to the second thing that stood out to me in Cursor’s press release:
Our in-house models now generate more code than almost any other LLM in the world.
While I am not convinced of that claimI’m convinced that the cursor is still moving despite its previous reliance on the foundation model. From Why some AI rappers create billion-dollar businesses again:
But cursors and other such tools are almost entirely dependent on access to Anthropic, OpenAI and Gemini models
open sourceOpen-weigh and in-house models match or surpass Frontier models in quality. developer forum Paying customers are full of complaints about rate caps. In my own projects, I exhausted my cloud credits at Cursor mid-project and, despite preferring Cursor’s user interface and design, I moved to cloud code (and had to pay ten times more to avoid the rate cap). The interface could be better, but model access proved decisive.
Cursor’s new in-house model Composer-2, which has just been launched The past month has been a good example of how this model versus application competition is evolving. Cursor claims (without any external benchmarks, I must say) that the Composer-2 is almost as good as the Frontier models, but 4 times faster. It is too early to say how true this is. Open-source models have not yet come close to the top spots in SWE-bench verified or private evaluations,
For me, model quality is the most decisive factor in these AI coding wars. And in my view, this is why Cloud Code has already overtaken Cursor, and OpenAI’s Codex is far behind, even though both were launched a year or so later.
Even though newcomers Cursor, Cloud Code, and OpenAI Codex (developers) remain the talk of the town, incumbents like Microsoft with GitHub Copilot, AWS with Kiro, and Google with AntiGravity can utilize their existing customer relationships, bundle their offerings with their existing suites, and/or provide their option as a default in their technology stack to compete. For example, Cursor charges $20-$40 monthly per user for productive use, while Google AntiGravity launched free with generous limits for individual users. Github Copilot still leads this market, proving once again that there are structural advantages in enterprise bundling and distribution. It’s the Classic Microsoft Teams vs. Slack Dynamic,
One way for startups to compete is to win over individual users who can use the coding tool with or without formal approval, and then become advocates for the tool inside the organization. That natural interest and adoption eventually compels IT and security teams to officially review the tool and then ultimately approve its use.
Yet, even as these new tools capture the developer mindshare, the underlying developer tools market is changing. Both IDE developers choose and resources They Our counseling has changed dramatically. StackOverflow, once the default for programmers stuck on a programming problem, has seen its traffic and question numbers soar. decline Dramatically since the launch of ChatGPT, it has been suggested that AI is already replacing some traditional developer resources.
Just as compilers freed programmers from writing assembly code, AI tools are freeing software engineers from the arduous work of writing boilerplate and routine code, and letting them focus on higher-order thinking. Ultimately, one day, AI may become so good that it will generate applications on demand and autonomously create entire software ecosystems. Both practical and applied AI coding tools, as well as incumbents and newcomers, see themselves as en route to fully autonomous software generation, even if they are taking different approaches. The ones that get there will be those that provide the best model quality that ship code reliably, go deep enough to ship features that foundation models can’t care enough to replicate, and become sticky enough that users won’t leave even when they could.,
If you enjoyed this post, please consider sharing it twitter/x Or LinkedinAnd tag me when you do.