--- title: "\"AI Programming Star\" Cursor CEO: The \"Third Era\" of AI Software Development Has Arrived" description: "Michael Truell, co-founder of Cursor, recently published an article dividing the evolution of AI programming into three stages: the era of Tab auto-completion, the era of synchronized agents, and the " type: "news" locale: "en" url: "https://longbridge.com/en/news/277343571.md" published_at: "2026-03-01T08:00:01.000Z" --- # "AI Programming Star" Cursor CEO: The "Third Era" of AI Software Development Has Arrived > Michael Truell, co-founder of Cursor, recently published an article dividing the evolution of AI programming into three stages: the era of Tab auto-completion, the era of synchronized agents, and the current "third era" dominated by cloud agents. In the new era, agents can operate independently, in parallel, and for extended periods, with developers transitioning from "code writers" to "agents' conductors." AI-assisted programming is undergoing a fundamental paradigm shift. Recently, Michael Truell, co-founder of Cursor, stated on the X platform that **the company has officially entered the "third era" of AI programming.** The core of this new era is **driven by cloud agents that can independently handle complex tasks and operate autonomously over long periods.** This means that Cursor's positioning has fundamentally shifted from "a tool for writing code" to "a platform that helps developers build software factories." Data is confirming this judgment: currently, **35% of the code submissions (PR) merged internally at Cursor are completed by autonomous agents running on cloud virtual machines.** More notably, **the company expects that within a year, the vast majority of software development work will be undertaken by such agents.** This trend will not only reshape the competitive landscape of AI programming tools but will also have a profound impact on the business models of the entire industry. ## From Tab to Agents: A Rapid Reversal of User Behavior The pace of evolution in AI programming is exceeding everyone's expectations. He reviewed the key turning points in this field and clearly divided the development of AI programming tools into three stages: **Stage One: The Tab Autocomplete Era.** The Tab key not only completes the current line but also intelligently predicts and completes the next line of code, even making multiple modifications across files, freeing developers from tedious coding. This stage lasted nearly two years, **with the core focus on automating low-entropy, repetitive tasks.** **Stage Two: The Synchronous Agent Era.** The core function of this era is conversational programming, where developers describe their needs to agents in natural language, and the agents generate code and respond in real-time, forming a rapid interactive loop of "prompt-feedback-correction." He predicts that this stage may last less than a year, with the transformation speed far exceeding previous expectations. **Stage Three: The Cloud Agent Era.** After developers hand over tasks, agents run independently in cloud virtual machines—autonomously completing coding, debugging, testing, and iteration—transforming developers from "code writers" to "agents' commanders." User behavior data confirms the intensity of this paradigm shift: in March 2025, the number of Tab users at Cursor was 2.5 times that of agent users; now, this ratio has completely reversed—agent users have reached twice the number of Tab users, and usage is still surging. He revealed that many Cursor users have completely stopped using the Tab key. ## Core Advantages of Cloud Agents: Parallelism and Asynchrony The limitation of synchronous agents lies in dual binding: they need to interact with developers in real-time while competing for computing resources with local machines. This means that the number of synchronous agents running simultaneously is extremely limited **The cloud-based intelligent agents fundamentally eliminate these two constraints.** Each agent runs on an independent cloud virtual machine, allowing developers to turn to other tasks after delivering the assignment, without needing to wait in real-time. Within a span of hours, the agents autonomously complete code iterations and testing validations, ultimately delivering results in the form of logs, recordings, and real-time previews, rather than displaying code differences line by line. **This delivery method makes it possible to run multiple agents in parallel.** Developers can quickly assess the output quality of multiple tasks without having to rebuild the context of each session from scratch. The role of humans thus undergoes a fundamental transformation: **from being the person who "guides the code line by line" to becoming the person who "defines problems and sets review standards."** ## Internal Practices Reveal a New Developer Work Model Cursor uses its own internal practices as an example to illustrate the specific form of this new work model. Developers adopting the new approach exhibit three common characteristics: **agents are responsible for writing nearly 100% of the code; developers focus their time on breaking down problems, reviewing products and code, and providing feedback; and multiple agents are launched simultaneously to advance in parallel, rather than guiding each one to completion hand-in-hand.** He also admitted that **this model still faces challenges for large-scale adoption across the industry.** At an industrial scale, the instability of tests or damaged environments that a single developer can bypass will evolve into systemic failures that interrupt each agent's operation. Additionally, ensuring that agents can access the complete tools and context they need remains a key issue to be resolved. Meanwhile, he stated that Cursor's recent release of new features is a "preliminary but important step" in this direction ### Related Stocks - [CLOU.US - Global X Cloud Computing ETF](https://longbridge.com/en/quote/CLOU.US.md) - [512720.CN - Guotai CSI Computer Theme ETF](https://longbridge.com/en/quote/512720.CN.md) - [159998.CN - Tianhong CSI Computer Theme ETF](https://longbridge.com/en/quote/159998.CN.md) - [IXN.US - ISHRS S&P Glb It](https://longbridge.com/en/quote/IXN.US.md) ## Related News & Research | Title | Description | URL | |-------|-------------|-----| | 08:37 ETCapxel Launches LLM-LD, the First Open Standard for Making Websites Readable by AI Agents | Capxel has launched LLM-LD, the first open standard designed to make website content readable by AI systems. This new sp | [Link](https://longbridge.com/en/news/276895952.md) | | Workday's AI Pivot Weighs on Near-Term Margins, Morgan Stanley Says | Workday's AI Pivot Weighs on Near-Term Margins, Morgan Stanley Says | [Link](https://longbridge.com/en/news/276913842.md) | | Why learning to 'speak AI' can help your money manager beat the market | The article discusses the importance of learning to 'speak AI' for finance professionals, emphasizing that proficiency i | [Link](https://longbridge.com/en/news/276920714.md) | | 22:53 ETArcfra Launches Neutree to Power Scalable Enterprise AI Infrastructure | Arcfra has launched Neutree, an enterprise-grade Model-as-a-Service (MaaS) platform aimed at enhancing AI operations. Th | [Link](https://longbridge.com/en/news/277138152.md) | | 12:02 ETLevel AI Announces Major AI Virtual Agent Expansion: Delivering Human-Quality AI for Every Interaction | Level AI is expanding its AI Virtual Agent capabilities within its customer experience platform, aiming to integrate hum | [Link](https://longbridge.com/en/news/277075464.md) | --- > **Disclaimer**: This article is for reference only and does not constitute any investment advice.