Skip to content
VibekollenBETAVibekollen

Source

Cursor

12 items in the feed. The texts are the sources' own descriptions — the content belongs to Cursor.

BlogCursor

How Grab put Cursor in the hands of Design, Ops, and Engineering

Grab, a Southeast Asian transportation service, has made Cursor available to almost everyone in the organization—not just engineers. Approximately 98 percent of employees in the tech organization use Cursor monthly. It has changed how work gets done: designers fix interface issues the same day instead of waiting for developers, engineers complete tasks in hours instead of days, and non-technical staff like finance teams and the CEO's office build their own tools. Over a third of code suggestions are accepted, and bug fixing is the top-ranked activity across all professional groups Grab studied.

15 Sept cursor.com

BlogCursor

Introducing Projects

Cursor is launching Projects, a tool that lets you assign a large work task to an AI coordinator — such as building a feature, executing a code migration, or maintaining an entire app. The coordinator maintains context over months, delegates work to thousands of subordinate AI agents, and can perform recurring tasks without you needing to request it each time. Cursor has been using Projects internally for several months for migrations, design system maintenance, and building Projects itself, and reports that new users merged 30 percent more PRs while experienced users merged six times as many.

10 Sept cursor.com

BlogCursor

Cursor Projects

Cursor is launching Projects, a new feature that enables work on larger code projects over extended periods. It includes a coordinator agent that plans work and delegates tasks to other agents without writing code itself. The project runs on a cloud server, allowing work to continue even when you close your computer. Agents share knowledge with each other, so if one agent learns something new about the codebase, all future agents can use that knowledge. You can also integrate Slack so the coordinator automatically starts work when needed—for example, when a bug is reported.

10 Sept cursor.com

BlogCursor

How Basis builds long-horizon accounting agents with Cursor

Basis builds AI agents for auditors that automatically perform complex accounting tasks over multiple hours — such as annual financial statements and tax declarations — and return results ready for review. The company uses Cursor to build and refine these agents, where they write instructions and rules in plain text that the agent follows. The main challenge is that long-running tasks contain hundreds of decisions where early mistakes can affect later steps, and a correct conclusion can hide an unreliable process. Basis solves this through "behavior specs" — Markdown documents that define how the agent should behave in specific situations, making it possible to assess the process, not just the result. An agent can complete work that normally takes 30-40 hours in approximately 6-7 hours, and Basis is reliable for 40 percent of the 25 largest audit firms.

4 Sept cursor.com

BlogCursor

Nokia analyzes 50M+ lines of code in two weeks with Cursor

Nokia is using Cursor, an AI tool for code analysis, to transform its entire software development. Two engineers analyzed over 50 million lines of code in two weeks — work that previously would have taken a dozen experts several months. Nokia is also using Cursor to build automation tools (one engineer created a project management system in one week that replaced 80 percent of the work for 6–10 project managers), speed up debugging from weeks to days, and generate product requirements from the start of the development process.

2 Sept cursor.com

BlogCursor

Run cloud agents on machines you manage

Cursor has launched Self-Hosted Machines, allowing AI agents to run on computers managed by your team instead of only in Cursor's cloud. You maintain control over the infrastructure and network, while Cursor still directs the agents' logic and planning. The feature is useful when agents need access to internal services, custom hardware like GPUs, or specific operating systems. Agents can now also run on elastic machine pools that automatically scale up when demand arrives and scale down after work is completed.

2 Sept cursor.com

BlogCursor

Self-hosted machines

Cursor has launched self-hosted machines, allowing you to run AI agents on your own infrastructure instead of in Cursor's cloud. Your code, build results, and secrets remain on your own computers. You can connect either a single laptop or virtual machine for personal use, or set up team pools — queues of workers that scale up and down as needed. The agent can now also see and control your computer screen on Linux and Mac, and can click, type, and navigate in web browsers.

2 Sept cursor.com

BlogCursor

Start from scratch, without a repo

Cursor, an AI-agent coding tool, now lets you start building without connecting an existing GitHub project or other code system. You write instructions directly to the agent, which creates a workspace for you. When satisfied, you can save the work as a real project in Cursor's own repository system. The tool also shows a live preview in the web browser, and you can publish your finished project directly via Vercel to get a public link.

27 Aug cursor.com

BlogCursor

IMDEX uses Cursor to build integrated subsurface data and analytics platform in months, not years

Mining technology company IMDEX uses Cursor, an AI coding tool, to build geological data platforms significantly faster than before. A project that would have taken two to three years and required 20 new engineers was completed in eight months. Another project, where they updated legacy code to modern technology, took six months instead of three to five years. IMDEX conducted intensive developer training over one month ("AI July") to teach staff how to effectively use AI tools in their daily work.

25 Aug cursor.com

← All news