Anyphere, the company that makes the assistant AI CODING cursor, is not thinking about an IPO anytime soon, the CEO of the founder CEO Michael Truell said On stage Monday at the Ai Fortune Brainstorming Conference.
After arriving $1 Billion in annual revenue in November, and raise $2.3 Billion at a cost of $29.3 billion Last month, Truell said the company was focusing more on building more features.
For example, he notes that the LLMS that has matured in the home already supports special products. Cursor also confirmed the shortcomings of the model in November when it was found, “The In-House model now generates more code than almost all other llms.”
Commenting on the model when the founders were asked how they plan to compete with LLM manufacturers they forgot when llm – openai, anthropopic – had its own AI coding offering.
Truell likes to code products for “concept cars” while his products are production cars.
“It’s going to be like taking the engine and the concept car around instead of the final production car,” Truell said. “What we do is we take the best intelligence for the market from the market from different providers. We also create product-specific models in our own places in the best places and improve and stop the UX to work with AI.”
The conditions of the cursor in its competitors – and the need to build its own LLMS – has been a topic of speculation in VCS in Silicon Valley since this year Openai is reported to be looking at anyone as an acquisition target. Anyphere became that idea. (This is at the same time Openai Windsurf trading also did not materializewith the founder eventually joining Google).
TechCrunch events
San Francisco
I’m fat
October 13-15, 2026
The problem, investors told Techcrunch, is AI coding editors lose money Thanks to the high fees paid to model manufacturers. In the case that the cursor, instead of selling, can set the price for the usage model in July, directly bypassing the cost of the API that creates the model for the user. The change from the all-inclusive subscription fee (and the huge bill that comes with it) is caused Uproar among some usersSee rank-.
On Monday, when asked about the price kerfuffle, Truell said, “When you started the cursor, you would create the cursor to get the hours worked for you and now you will model it for you and everything else in the consumption model,” he said.
Truell added that one of the tools the Company uses is a cloud-compatible tool, which can create a tab of total usage and keep it on a running bill.
“We have a whole team dedicated to enterprise engineering and building things like spend control and group billing and visibility,” he said.
Additionally, he said the cursor has focused on two key areas for next year. One is to handle more complex agent functions.
“We want you to take the last task, which is brief to define but then it’s really difficult to do, and it has to be done with the cursor. For example is bug fix,” he explained.
He wanted the cursor to fix the various bugs he might have described, but it took “weeks of time, thousands of code runs” to solve. “We want the cursor to do that, end-to-end,” he said.
Another area he named, but didn’t explain in detail, was the idea of ”thinking about the team as a given atomic unit,” he said. This should be different to serve the individual codter, and guide well in the company.
In addition to the wastage feature, Truell said he wanted the cursor to address the part of the software development lifecycle outside of writing code. He pointed to the Review Cursor product as an example, which he said is used by some customers to analyze each customer to analyze each request, whether written by AI or human. (A pull request is when a programmer submits code for review before it is merged into the main project).
“So you’ll see us starting to help the team more,” with more features like that, he promised.
Meanwhile, the big competitors are also all fierce for the world of complex-tasking agents. Amazon only publishes the promised code tools has been open for days.
Just this week, AI power players, including anthropopic, openai, Microsoft, AWS and Many otherslaunched a new consortium at the Linux Foundation to develop agent agent interoperability standards. He even presented some major projects, such as the popular context model “anthropolMcp).
The plan for the year is likely not to be placed in front of the main model competitors that include the Cursor. However, it should remain the company.