From 0 to 1: My Hands-on Experience Developing a Web App Entirely with Cursor

Author: Liang Bo Eric
Original Link: Original Article

---

Greetings, let me share my practical experience and reflections.

Meta CEO Mark Zuckerberg stated: "By 2025, software engineering will undergo dramatic changes, with AI reaching the programming level of intermediate software engineers." He also plans to delegate much application programming work to AI. As AI capabilities continue to advance, many are understandably concerned about job displacement.

After recently developing a web application entirely through natural language using Cursor + Claude 3.7, I'd like to share some perspectives on AI-assisted programming.

01 Embrace AI, Acknowledge Its Power

Since ChatGPT's launch in 2022, AI has entered mainstream awareness. Gradually becoming an indispensable tool, discussions about AI replacing various professions have emerged. When such predictions target one's own industry, most people instinctively resist, dismissing them as alarmist.

Take programmers for example - current large models like Claude 3.7 already demonstrate coding capabilities comparable to mid-senior developers. By leveraging AI effectively, junior developers can achieve what previously required intermediate-level skills, potentially surpassing them.

02 The Growing Importance of Requirements Analysis

The ability to analyze, understand, and break down requirements is becoming increasingly crucial. Imagine working with AI employees - clear instructions and systematic task decomposition for complex projects enable AI to complete work step-by-step. This mirrors how team leads manage development teams in actual projects.

For instance, when developing a website with Cursor (an AI-powered editor), simply instructing "Develop a website using React" yields only a basic demo. However, by systematically organizing specific requirements in Notepad, implementing them incrementally with proper specifications, the AI can gradually produce code that meets your needs.

03 Integrating AI into Personal Workflows

Identify aspects of your professional and personal workflows where AI can assist. For full-stack developers, AI can help generate UI mockups, assist with coding, or analyze source code. Even within AI-assisted programming, there's significant difference between Copilot-style assistance and full natural language programming.

Deep integration of AI into personal workflows should deliver tangible efficiency gains and quality improvements, not just mindless adoption of AI-generated outputs.

04 Will AI Take Your Job?

While remarkably capable today, AI still struggles with many complex real-world projects. Developing an application from scratch using only natural language reveals AI's dual role as both mentor and employee. Its vast knowledge often produces impressive code, yet it frequently requires iterative guidance for proper implementation.

For simpler scenarios with established patterns, AI can indeed outperform many humans.

Rather than stealing jobs, AI is transforming roles and demanding new skills in AI collaboration.

Appendix

Source Code:

Nearly all code was written using Cursor + Claude 3.7.

https://github.com/flyeric0212/wx-md [1]

Development Process:

Requirements were decomposed using Notepad, with code standards enforced via .cursorrules