NVIDIA 已成为人工智能的代名词,也是全球市值最高的企业。去年,该组织设定了一项新的工程任务:利用 Cursor 将 AI 嵌入到软件开发生命周期的每一个阶段,并消除代码生成、测试、审查和调试环节中的人工瓶颈。
如今,超过 3 万名开发者每天使用 Cursor,使得提交的代码量增加了三倍以上。除了代码生成,NVIDIA 还针对其工程工作流对 Cursor 进行了定制,将 AI 模型的影响力从提升个人生产力扩展到自动化软件开发生命周期中的核心生产工作流。
Cursor 在大型代码库上表现更佳
在其 30 年的历史中,NVIDIA 积累了包含多种技术栈的大型代码库。这些代码库通过许多共享依赖项紧密交织在一起。一个代码库的变更常常会对其他代码库产生下游影响,这使得即使是最优秀的工程团队也难以驾驭这个复杂系统的细微之处。
NVIDIA 的每条产品线都有一个快速演进的复杂代码库。开发者很难跟上这些变化并理解整个代码库。这正是 Cursor 真正大放异彩的地方。
Wei Luo
NVIDIA 工程副总裁
NVIDIA 发现 Cursor 在其环境中能快速、准确地给出结果。这种差异源于 Cursor 能够对大型代码库进行映射和语义推理。高级软件架构师 Fabian Theuring 解释说,Cursor 的智能体明显更智能、更快速、更高效,因为它只检索最相关的上下文。
Cursor 在 NVIDIA 开发环境中的速度和准确性,立即对工程效率产生了影响。
在 Cursor 之前,NVIDIA 使用过其他 AI 编码工具,既有内部自建的,也有来自其他外部供应商的。但在采用 Cursor 之后,我们才真正开始看到开发效率的显著提升。
Wei Luo
NVIDIA 工程副总裁
从代码生成到软件开发生命周期的端到端自动化
随着 NVIDIA 的开发者借助 AI 更快地交付代码,瓶颈转移到了软件开发生命周期的其他阶段:代码审查、测试和调试。NVIDIA 的工程领导层设定了雄心勃勃的目标,将 Cursor 也扩展到这些工作流程中。“我的使命是将 AI 嵌入到软件开发生命周期的每一个环节,”Luo 说道。
Cursor 几乎被用于所有产品领域以及软件开发的各个方面。各个团队都在使用 Cursor 编写代码、进行代码审查、生成测试用例以及质量保证。我们的整个软件开发生命周期都因 Cursor 而加速。
魏罗
NVIDIA 工程副总裁
这一切始于将 Cursor 的用例从代码生成扩展到调试等领域。Theuring 解释说:“Cursor 在发现和解决罕见、持久的错误方面表现出色。”Cursor 不仅能持续识别这些问题,还能派遣智能体去解决它们,这一点尤其影响深远。
NVIDIA 还配置了 Cursor 来自动化整个工作流程。例如,Theuring 的团队正在使用自定义规则来自动化 Git 流程:分支创建、代码提交、持续集成调试和问题追踪。Luo 的团队则采用类似的方法进行错误修复,其自动化流程从使用 MCP 服务器从工单和文档中提取上下文开始,最后以 Cursor 实施错误修复并运行测试进行验证结束。Cursor 的可扩展性将其影响范围从个人生产力扩展到了项目层面。
我们在 Cursor 中构建了大量自定义规则,以完全自动化整个工作流程。这释放了 Cursor 的真正潜力。
Fabian Theuring
NVIDIA 高级软件架构师
更快的上手时间与压缩的学习曲线
Cursor 还帮助 NVIDIA 的新员工更快地熟悉不熟悉的代码库,并在比以往更短的时间内开始做出贡献。
它也让资深开发者能够在新编程语言或技术栈的各个部分迎接新的挑战。例如,经验丰富的后端工程师现在比以前更有信心处理前端任务。“Cursor 让开发者能够弥合技能差距,并在新领域更快上手,”Luo 解释道。
衡量开发速度和质量的综合价值
NVIDIA 正在通过几个关键指标来衡量 Cursor 的影响:
- 采用率:超过 3 万名开发者每天使用 Cursor
- 编码速度:使用 Cursor 的开发者提交代码量是此前的三倍
- 代码质量:尽管编码速度提升,Bug 率仍保持平稳,且代码风格一致性有所改善
我们每天都在使用 Cursor,现在已经回不去了,因为它彻底改变了软件工程的工作方式。构建软件比过去有趣得多。我真的很喜欢它。
法比安·托伊林
NVIDIA 高级软件架构师
如果您对构建 AI 原生工程团队感兴趣,请与我们的团队联系,开始 Cursor 试用。
NVIDIA has become synonymous with AI and the most valuable enterprise in the world. Last year, the organization set a new engineering mandate: leverage Cursor to embed AI across every phase of the software development lifecycle (SDLC) and eliminate manual bottlenecks across code generation, testing, reviews, and debugging.
Today, over 30,000 developers use Cursor daily, driving an over three-fold increase in committed code. Beyond code generation, NVIDIA customized Cursor for its engineering workflows, extending the impact of AI models from individual productivity enhancements to the automation of core production workflows across the SDLC.
Cursor delivers better performance on large codebases
Over its 30 year history, NVIDIA has amassed massive codebases with varied tech stacks. These codebases are closely intertwined with many shared dependencies. Changes in one codebase often have downstream effects on others, making it difficult for even the best engineering teams to navigate the nuances of this complex system.
Each of NVIDIA's product lines has a complex codebase that is evolving quickly. It's very hard for developers to stay on top of these changes and understand the entirety of the codebase. This is where Cursor really shines.
Wei Luo
VP of Engineering, NVIDIA
NVIDIA saw fast, accurate results with Cursor in its environment. This difference is due to Cursor's ability to map out and semantically reason over large codebases. Fabian Theuring, a senior software architect, explained that Cursor's agent is noticeably smarter, faster, and more efficient because it retrieves only the most relevant context.
Cursor's speed and accuracy across NVIDIA's development environment made an immediate impact on engineering velocity.
Before Cursor, NVIDIA had other AI coding tools, both internally built and other external vendors. But after adopting Cursor is when we really started seeing significant increases in development velocity.
Wei Luo
VP of Engineering, NVIDIA
From code generation to end-to-end automation of the SDLC
As NVIDIA's developers began shipping code faster with AI, bottlenecks shifted to other phases of the SDLC: code review, testing, and debugging. NVIDIA's engineering leadership set ambitious goals to extend Cursor into these workflows as well. "My mission here is to embed AI in every step of the SDLC," says Luo.
Cursor is used in pretty much all product areas and in all aspects of software development. Teams are using Cursor for writing code, code reviews, generating test cases, and QA. Our full SDLC is accelerated by Cursor.
Wei Luo
VP of Engineering, NVIDIA
It started with expanding the use cases for Cursor beyond code generation to areas like debugging. Theuring explained that "Cursor excels at finding and resolving rare, persistent bugs." Cursor's ability to not only consistently identify these issues but also dispatch agents to solve them has been particularly impactful.
NVIDIA has also configured Cursor to automate entire workflows. For example, Theuring's team is using custom rules to automate the git flow: branch creation, code commits, CI debugging, and issue tracking. Luo's team is taking a similar approach to bug fixes with automation that starts by pulling context from tickets and documentation using MCP servers and finishes with Cursor implementing bug fixes and running tests for validation. Cursor's extensibility expanded the scope of impact from individual productivity to program level impact.
We have built a lot of custom rules in Cursor to fully automate entire workflows. That has unlocked Cursor's true potential.
Fabian Theuring
Senior Software Architect, NVIDIA
Faster ramp times and compressed learning curves
Cursor is also helping NVIDIA's new hires get up to speed on unfamiliar codebases and start contributing in a much shorter timeframe than before.
It has also allowed senior developers to take on new challenges across new programming languages or parts of the tech stack. For example, experienced backend engineers are tackling frontend tasks more confidently than before. "Cursor allows developers to bridge their skill gaps and ramp in new areas faster," explained Luo.
Measuring value across development velocity and quality
NVIDIA is measuring Cursor's impact across a few key metrics:
- Adoption: Over 30,000 developers use Cursor daily
- Coding velocity: Developers using Cursor commit three times more code than before
- Code quality: Bug rates have stayed flat despite increases in coding velocity, and consistency in code style has improved
We are using Cursor every day, and now there's no going back because it has completely changed the way software engineering works. Building software is now a lot more fun than it used to be. I really love it.
Fabian Theuring
Senior Software Architect, NVIDIA
If you're excited about building AI-native engineering teams, please reach out to our team to get started with a Cursor trial.