Opinion AI-integrated development environment (IDE) company Cursor recently [1]implied it had built a working web browser almost entirely with its AI agents. I won't say they lied, but CEO Michael Truell certainly [2]tweeted : "We built a browser with GPT-5.2 in Cursor."

He followed up with: "It's 3M+ lines of code across thousands of files. The rendering engine is from-scratch in Rust with HTML parsing, CSS cascade, layout, text shaping, paint, and a custom JS VM."

That sounds impressive, doesn't it? He also added: "It *kind of* works," which is not the most ringing endorsement. Still, numerous news sources and social media chatterboxes ran with the news that [3]AI built a web browser in a week .

[4]

Too bad it wasn't true. If you actually looked at Cursor engineer Wilson Lin's blog post about [5]FastRender , the AI-created web browser, you won't see much boasting about a working web browser. Instead, there's a video of a web browser sort of working, and a much less positive note that "building a browser from scratch is extremely difficult."

[6]

[7]

The thing about making such a software announcement on GitHub is that while the headlines are proclaiming another AI victory, developers have this nasty trick. They actually git the code and try it out.

[8]Developers quickly discovered the "browser" barely compiles , often does not run, and was heavily misrepresented in marketing.

[9]

As a techie, the actual blog post about how they tried and didn't really succeed was much more interesting. Of course, that Cursor sicced hundreds of GPT-5.2-style agents which ran for a week to produce three million lines of new code, to produce, at best, a semi-functional web browser from scratch, doesn't make for a good headline.

According to Perplexity, my AI chatbot of choice, this week‑long autonomous browser experiment consumed in the order of 10-20 trillion tokens and would have cost several million dollars at then‑current list prices for frontier models.

I'd just cloned a copy of Chromium myself, and for all that time and money, independent developers who cloned the repo reported that the codebase is very far from a functional browser. Recent commits do not compile cleanly, GitHub Actions runs on main are failing, and reviewers could not find a single recent commit that was built without errors.

[10]

Where builds succeeded after manual patching, performance was abysmal, with reports of pages taking around a minute to load and a heavy reliance on existing projects like [11]Servo , a Rust-based web rendering engine, and QuickJS, a JavaScript engine, despite "from scratch" claims.

Lin defended the project on Y Combinator, [12]saying , for instance: "The JS engine used a custom JS VM being developed in vendor/ecma-rs as part of the browser, which is a copy of my personal JS parser project vendored to make it easier to commit to." If it's derived from his personal JavaScript parser, that's not really from scratch, is it? Nor is it, from the sound of the argument, written by AI.

[13]Just because Linus Torvalds vibe codes doesn't mean it's a good idea

[14]The Microsoft 365 Copilot app rebrand was bad, but there are far worse offenders

[15]The most durable tech is boring, old, and everywhere

[16]What the Linux desktop really needs to challenge Windows

Gregory Terzian, a Servo maintainer, [17]responded : "The actual code is worse; I can only describe it as a tangle of spaghetti... I can't make much, if anything, out of it." He then gave the backhanded compliment: "So I agree this isn't just wiring up of dependencies, and neither is it copied from existing implementations: it's a uniquely bad design that could never support anything resembling a real-world web engine." Now that's a burn.

From where I sit, what makes the Cursor case more dangerous than just a failed hack‑week project is that the hype is baked into its methodology. The "experiment" wasn't presented as what it really was: an interesting, but messy, internal learning exercise. No, it was rolled out as a milestone that conveniently confirmed the company's long‑running autonomous agent advertising. Missing from the story were basics any senior engineer would demand: passing Continuous Integration (CI), reproducible builds, and real benchmarks that show the browser doing more than limping through a hello-world page.

Zoom out, and CEOs are still predicting that AI will write 90 percent of code in a year, while most enterprise AI pilots still fail to deliver meaningful return on investment.

We're now in a kind of AI uncanny valley for developers. Sure, tools like Cursor can be genuinely helpful as glorified autocomplete and refactoring assistants, but marketing keeps insisting junior engineers can take whole projects from spec to shipping. When you start believing your own sizzle reel, you stop doing the tedious validation work that separates a demo from a deliverable.

Enough already. The hype has grown cold. Sarah Friar, OpenAI's CFO, recently blogged that in 2026, its focus would be on " [18]practical adoption ." Let's see real-world practical results first, and then we can talk about practical AI adoption. ®

Get our [19]Tech Resources



[1] https://cursor.com/blog/scaling-agents

[2] https://x.com/mntruell/status/2011562190286045552

[3] https://medium.com/write-a-catalyst/ai-agents-built-a-web-browser-in-one-week-and-that-should-make-us-pause-1be4fab67d03

[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/agenticai&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aXedyQikQXIQDYnSZ2Ck_gAAAQ4&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0

[5] https://github.com/wilsonzlin/fastrender

[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/agenticai&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aXedyQikQXIQDYnSZ2Ck_gAAAQ4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/agenticai&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aXedyQikQXIQDYnSZ2Ck_gAAAQ4&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[8] https://emsh.cat/cursor-implied-success-without-evidence/

[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/agenticai&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aXedyQikQXIQDYnSZ2Ck_gAAAQ4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0

[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_specialfeatures/agenticai&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aXedyQikQXIQDYnSZ2Ck_gAAAQ4&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0

[11] https://servo.org/

[12] https://news.ycombinator.com/item?id=46650998

[13] https://www.theregister.com/2026/01/16/linus_torvalds_vibe_coding/

[14] https://www.theregister.com/2026/01/09/microsoft_365_copilot_app/

[15] https://www.theregister.com/2025/12/31/long_lived_tech/

[16] https://www.theregister.com/2025/12/22/what_linux_desktop_really_needs/

[17] https://news.ycombinator.com/item?id=46709191

[18] https://www.theregister.com/2026/01/20/openai_money/

[19] https://whitepapers.theregister.com/