Cursor is in discussions for a funding round that could value the company at $50 billion. This would nearly double its $29.3 billion valuation from November when it raised $2.3 billion in a round co-led by Accel and Coatue. The startup attracts backing from investors like Coatue, Thrive Capital, and Andreessen Horowitz, as well as major technology companies like Google and Nvidia. While community members in 2024 suggested the startup died because its cloud backend went offline or its developer team shrunk to two people, the company maintains 1 million daily active users and 50,000 businesses. Its annualized revenue surpassed $2 billion in February, and 60% of revenue comes from enterprise customers like OpenAI and Budweiser. The revenue run rate doubled compared with three months earlier. The platform, which launched in 2023, popularized "vibe coding," where developers use simple prompts to generate applications. Cursor users can switch between models from Google, Anthropic, and OpenAI. One co-founder departed in late 2023, and some users noted the interface changed too much in a single year.
GitHub Copilot Enterprise and its coding agent
GitHub Copilot Enterprise provides customization and integration with an organization’s codebase. This plan costs an extra $20 per user per month over the $19 Business plan. The tool includes a coding agent that was introduced at Microsoft Build. This agent spins up a secure development environment, pushes commits to a draft pull request, and allows users to track progress through session logs. The coding agent assists with application modernization, including refactoring and rewriting code, but requires human approval before any CI/CD workflows run. GitHub Copilot also includes Claude Sonnet 5, Claude Opus 4.8 fast mode, and the MAI-Code-1-Flash model. The tool integrates Bing search in a beta capacity and allows developers to use accessibility functions to start sessions by saying "Hey code." Enterprises can also use branch rulesets to block pull requests if test coverage drops below set thresholds. The agent starts its work when a user assigns a GitHub issue to it or prompts it in VS Code, helping junior developers contribute quicker and senior developers handle live incidents. Organizations can also create custom models and manage dependency licenses at scale. Enterprises can also train Copilot on private codebases to provide more relevant suggestions, but the tool will not use private repositories or prompts to train its underlying machine learning models unless the company expressly directs it.
Comparing accuracy and costs
I find that the choice between these tools depends entirely on whether you prioritize a polished IDE experience or broad language support. Cursor costs $15 per month for Pro users, while GitHub Copilot Pro costs $10 per month. Cursor offers higher tiers like Pro+ at $60 and Ultra at $200, which provide 3x and 20x usage respectively. Cursor provides support for 10 languages, whereas GitHub Copilot provides support for over 50 languages, including Ruby and Go. While Cursor has a free tier for basic use, GitHub Copilot has no free tier. While Cursor is helpful for beginners, it struggles with complex tasks and larger projects that require precise code, whereas GitHub Copilot excels at complex queries and rapid code snippets. GitHub Copilot can generate incorrect or insecure code, so you must review its outputs carefully. For those checking test quality, the SLOBAC audit is a Cursor Agent Skill that can take about 30 minutes to run on large suites. For context, Windsurf provides 50K to 70K tokens, Cursor defaults to 200K, and Claude Code offers 1M tokens. For debugging, Cursor provides basic assistance while GitHub Copilot provides none. For developers who want to work with various models, Cursor allows switching between models from Google, Anthropic, and OpenAI. GitHub Copilot remains a strong option for experienced developers who require rapid prototyping and advanced features. You should test the tools on your own codebase. How can a developer ensure an AI does not introduce security vulnerabilities during an automated refactoring session?
