not really interested in sharing my project details tbh "Cursor orchestrated hundreds of AI agents using GPT-5.2-Codex to build a basic but functional web browser from scratch. The AI agents generated over 3 million lines of code in just one week - project called FastRender"

Two data points dropped in the last few months that should terrify every software company that thinks its codebase is a moat. First, one engineer at Cloudflare, working with Claude via AI agents, rebuilt 94% of Next.js, one of the most widely used frontend frameworks on the internet, built over 10 years by a large engineering team in a single week. Total cost was $1,100 in API tokens. The result, called Vinext, is a drop-in replacement that builds production apps up to 4x faster and produces client bundles 57% smaller and customers are already running it in production. Second is Cursor CEO Michael Truell deployed a swarm of hundreds of GPT-5.2 agents that ran uninterrupted for an entire week and built a fully functional web browser from scratch called FastRender. 3 million lines of code, thousands of files and a custom Rust rendering engine with HTML parsing, CSS layout, text shaping, and a custom JavaScript VM. Total cost was roughly $30,000. For context, Google has spent billions of dollars and decades of engineering building Chrome. And the benchmarks say by next year, you will be able to one-shot prompt anything. The moat that software companies spent decades building, the complexity of their codebase, the years it would take a competitor to replicate it, the switching costs that moat assumed humans were the unit of production. AI does not care how long it took you to build it, it only cares how long it takes to rebuild it. And right now, the answer is one week.

TIL that the cursor browser project (fastrender) was by the same person (wilson lin) who wrote this amazing piece last year

Flow-Swarm @OpenAI Codex, running autonomously for 25 hours straight. My own take on @cursor_ai's FastRender experiment. Obviously I can't run 2000 agents for a week. Had to scope it way down. Bought two additional fresh ChatGPT Pro accounts just for this. Peak concurrency: 10-11 agents (beginner numbers but still). Orchestrating GPT-5.3 Codex (high) (@OpenAIDevs cooked) has been impressive. Stays on task for hours. Doesn't drift. It's just a phenomenal model. Not going to say what it is yet. But there's an entire category of software that's been coasting for years because building a good version is (was!) genuinely hard. So everyone ships something half-broken and charges enterprise prices for it. It's about 5% away from acceptance gates I defined right now. The incumbents should be nervous.

約莫一個月前 Cursor 發表了一篇文章宣稱用 AI 寫了瀏覽器 我花了一點時間研究滿百感交集的 因為他其實只是完全抄襲 Servo 的程式碼 而對我來說他抄襲了我所寫的所有程式碼 但我並不是感到生氣 畢竟整個 fastrender 其實一團亂根本不可能當作真正的瀏覽器 (1/2)

Codex 5.3 和 Opus 4.6 确实很强，但是有多强，目前我试了几个项目，感觉前一代也能解决，现在想榨干模型的能力，几乎是不可能的。甚至想找一些前一代模型搞不定、但新模型却能轻松搞定的案例都很难。 可以说，LLM 现在是 OverSkill 的 看到一个 blog 说，目前关于新模型能力最有说服力的故事，是 Anthropic 的 Nicholas Carlini 谈 Opus 4.6 的那篇：用一支并行协作的 Claude 团队来构建一个 C 编译器——这可以看作是 Anthropic 版的 Cursor “FastRender”项目。 anthropic.com/engineering/bu…

Миллион строк кода и ни одного рабочего билда: мифы автономного кодинга от Cursor Громкий анонс браузера, созданного ИИ-агентами, обещал революцию. Однако при ближайшем рассмотрении репозиторий Fastrender оказывается набором «ИИ-слопа»: u.habr.com/bJYLt

I loved this end: “FastRender is effectively using a full browser rendering engine as a “hello world” exercise for multi-agent coordination” with massively parallel coding agents, the bar from “Hello World” is shifting from 1 line to a million lines!

创意AI Agent实验引发关注：基于Qwen 2.5的纯客户端AI尝试通关宝可梦红，一个'纯AI'社交网络上线让模型自主发帖、吵架、交友，FastRender展示并行Agent造浏览器。OpenAI也公开了Codex Agent循环架构细节。

Creative AI agent experiments are capturing attention: A fully client-side AI using Qwen 2.5 attempts to beat Pokemon Red, an 'AI-only' social network where models autonomously post, argue and form relationships launched, and FastRender demonstrates parallel agents building a browser. OpenAI also detailed its Codex agent loop architecture. #AIAgents #GamingAI #SocialNetwork #Codex

Were those new features in Cursor directly influenced by the FastRender work? That is an impressively quick turnaround from research to production!

I had a fascinating conversation with @wilsonzlin about FastRender, the browser rendering engine he built with the help of 2,000 coding agents over the past few weeks. It's 47m on YouTube or you can read my highlights here: simonwillison.net/2026/Jan/2…

We have two existence proofs of how long it would take to build a new web renderer by hand – Servo and Ladybird. Both have consumed dozens of person-years of effort by some of the best engineers on earth. Now clearly they are both higher quality than fastrender. But still.

ecma-rs is by the same author as FastRender taffy is the one dependency which did raise my eyebrow though - they document changes they made to that here but it does feel like a pretty key component for them to outsource github.com/wilsonzlin/fastre…

Cursor used hundreds of autonomous agents that collaboratively built a functional Rust-based web browser, FastRender, from scratch in under a week. The project generated over a million lines of code across 1,000 files, using a hierarchical agent system.

под названием FastRender наглядно показал, что ИИ уже способен брать на себя сложные задачи по программированию без участия человека, сокращая месяцы и даже годы разработки до считанных дней. hightech.plus/2026/01/16/sot…

I went through Cursors fastrender repository. A lot useful stuff how they did their multi-agent workflow.