AI writes 3 Million lines of code and builds a full web browser in just one week, internet asks does it work

By Admin

20 Jan, 2026

Introduction
Artificial intelligence is evolving at a pace that continues to surprise developers, businesses, and the global tech community. From generating content to writing complex software, AI-driven coding is no longer a futuristic concept. A recent revelation has taken this conversation to the next level: an AI system reportedly wrote more than 3 million lines of code in just one week to build a complete web browser. While the achievement sounds extraordinary, the internet remains divided on one critical question does it actually work?

AI Builds a Web Browser in Seven Days
The discussion began after Michael Trulli, CEO of Cursor, shared details of an internal experiment on X. According to Trulli, the company tasked an AI coding agent with building a full web browser from scratch. The result was staggering: over 3 million lines of code generated across thousands of files in just seven days, without any direct human intervention.

The AI model used for this ambitious experiment was GPT 5.2 Codex, developed by OpenAI specifically for extended autonomous coding tasks. The browser was written largely in Rust and included core components such as HTML parsing, CSS cascading, layout systems, text shaping, painting, and even a custom JavaScript virtual machine.

Does the AI-Made Browser Actually Work?
To support his claims, Trulli shared a screenshot showing the browser loading the Google homepage. However, he was quick to manage expectations. In his own words, the browser “kind of works” but still faces significant limitations. It is far from matching the performance, stability, and feature completeness of established engines.

Trulli openly acknowledged that the project is nowhere near production-ready and should be viewed purely as an experimental milestone. He clarified that the goal was not to replace existing browsers, but to explore how far autonomous AI coding agents can be pushed in a short time frame.

Internet Skepticism and Comparisons
Despite the impressive numbers, internet users responded with healthy skepticism. Many pointed out that Chromium, which powers Google Chrome and Microsoft Edge, contains more than 35 million lines of code developed over many years by thousands of contributors.

Some users questioned the prompts used to guide the AI, while others wondered how many Cursor tokens such a task would consume for a regular user. There were also debates about whether the AI was instructed to take architectural inspiration from Chromium or other existing browser engines.

Not an Isolated Case in AI Autonomy
This experiment is not an isolated incident. AI systems are increasingly demonstrating the ability to work autonomously for extended periods. Recently, Anthropic claimed that its model Claude Sonnet 4.5 can code a complete application in approximately 30 hours, further fueling discussions around AI-driven software development.

What This Means for the Future of Coding
While the AI-built browser may not be ready for everyday use, the implications are significant. Writing millions of lines of functional code in a week was once unimaginable. Experiments like this highlight how “vibe coding” and autonomous coding agents could transform software development, reduce development timelines, and reshape how engineers collaborate with AI.

At the same time, the skepticism is justified. Code quality, security, maintainability, and real-world usability remain critical challenges. For now, AI-built browsers may be experimental, but they clearly signal a future where AI plays a central role in building complex digital infrastructure.

Conclusion
The idea of an AI writing 3 million lines of code and creating a web browser in just one week is both exciting and unsettling. While the project does not yet rival established browsers, it represents a major step forward in autonomous AI coding. Whether this technology becomes a foundation for future production systems or remains an experimental benchmark, one thing is certain: the boundaries of what AI can build are expanding faster than ever.