The New Browser, Same as the Old
Lately, it feels like every day brings another headline about AI reshaping how we live, work, or interact online. Most of it blends together. Press releases dressed up as journalism, written in that same old tired, strained language of innovation, like it’s trying too hard to sound new.
But one of these headlines recently caught my attention:
“OpenAI to release web browser in challenge to Google Chrome.”
I clicked. It read like the usual fluff: strategic moves, product rollout, innovation, competition.
Then, buried in the middle:
“OpenAI’s browser is built atop Chromium.”
That sentence said everything.
This alleged challenger to Chrome runs on the same engine that makes Chrome Chrome.
OpenAI claims this new browser will change how we interact with the web. By bolting their AI agents onto a Chromium base, they aim to “directly integrate AI agents into the browsing experience, enabling the browser to carry out tasks on behalf of the user.” But they’re doing it while reinforcing the very framework that already defines how we use the web today.
This is just another in a long line of tools that promise innovation while quietly inheriting the flaws and assumptions of the systems they claim to replace.
New interfaces built on old infrastructure. New language around old power.
So let’s talk about this one.
I want to show why OpenAI’s decision was both predictable and still a mistake. And what I think their real reason was for choosing Chromium. From there, we’ll look at why so many so-called “alternatives” end up caving into compliance, why almost no one builds from scratch anymore, and what a true alternative might have looked like if someone had the will, and the computing power, to build it.
New Chrome Finish
It seems every new browser is Chromium now.
Brave. Opera. Arc. Even Microsoft Edge.
And now, OpenAI’s joining the ranks.
There’s a good reason.
The Chromium engine became the standard because it worked. It was fast. It was stable. Developers liked it. Users liked it.
But over time, that convenience has turned into control. Chrome became the default, so web developers built for Chrome.
And now, anything built on Chromium inherits that same expectation. Sites load. Buttons behave. Web apps just work.
But here’s the part no one likes to mention:
Every time another browser adopts Chromium, Google’s grip on the web tightens.
Google dictates the standards. Developers follow them.
If your site doesn’t play by Google’s rules, it might not load properly.
Sorry, Firefox. Sorry, Safari. Google runs the rendering game now.
It doesn’t matter if you block ads.
It doesn’t matter if you swap out the search engine.
If your browser is built on Chromium, you’re reinforcing Google’s blueprint for how the web should work.
You’re not breaking their monopoly.
You’re laundering it.
A fresh coat of branding. A new name on the tab.
But underneath? The same plumbing. The same rules.
This is what laundering looks like in tech: taking a dominant system, rewrapping it, and selling it back to users as if it’s something independent, neutral, even rebellious.
It’s monopoly by proxy. Power wearing a mask.
So when OpenAI, a company supposedly positioned as Google’s rival, builds its browser on Chromium, it doesn't disrupt the system.
It validates it.
They’re not rewriting the rules.
They’re just making them look different.
Because the real game isn’t browsers.
It’s data, and control over who sees what, when.
Enhanced Personalization
Every major browser except Brave shares users’ browsing history and unique identifiers with the companies that own them. This data is often stored on corporate servers, where it can be shared with third parties, handed over to governments, or exposed in a breach.
Add an AI agent to that mix, and it gets worse.
They’re being marketed as assistants, but they’re behavior-mapping tools.
They learn your patterns. Your voice. Your impulses.
They remember your grocery list, your late-night questions, the way you phrase things when you're tired, or scared, or angry.
And now, many people are using AI for everything, including search.
Why go to Google when ChatGPT gives you better answers?
This isn’t hypothetical.
It’s already happening.
Google sees it. That’s why they’re panicking.
OpenAI sees it too. And now they want the browser itself.
Not the page. Not the link. The entire interface.
A chatbot between you and the internet.
One that decides what you see, and remembers how you asked to see it.
And the real reason OpenAI chose Chromium? The license.
Chromium’s permissive terms let them build more privately, without the transparency requirements others would demand.
What that means in practice, we don’t yet know.
But if history is any guide, we won’t get to inspect what it remembers.
We won’t get to audit what it knows.
We’ll just type.
And hope.
Fork in the Engine
Yet hope isn’t the only option. Just the easiest one.
For all its risks, there was another path: Gecko, the rendering engine behind Firefox.
Not as shiny. Not as easy.
More bugs. More work. More risk.
But also: independence.
A different lineage. A different vision.
A real alternative.
Gecko isn’t dead, but it’s been limping for years.
Google’s dominance over web standards means devs optimize for Blink first, and everything else later.
WebKit might seem like an option, but it answers to Apple’s walled garden, not the open web. Gecko is the last general-purpose engine with real independence.
And with OpenAI’s 500 million weekly users, it could’ve been viable again.
It might’ve sparked the first real break in the browser monoculture in years.
OpenAI had the money. They had the talent. They just didn’t want to.
Because this isn’t about building a better browser. It’s about building a better way to harvest data. At least, that’s how it appears. Maybe it was also about ease of integration, or scalability. But perception matters.
The Browser That Could Be
Someone could still make an AI browser work while maintaining user privacy.
Start with a rendering engine that doesn’t report back to Mountain View. Pair it with one of the many open-source language models we already have: DeepSeek, Mistral, take your pick.
Build a browser with these agents you control.
It could be private by default. Not as a toggle buried in “Advanced Settings.”
Built that way from the ground up.
No data harvesting.
No shadow memory.
No automatic handoff to a cloud you can’t see.
Just a browser.
With an integrated assistant that doesn’t need to know everything about you to be useful. With memory stored locally, not in the cloud.
This is all possible, right now. Every component already exists.
The FLOP Reality
The barrier isn’t the technology. It’s economics.
A private AI browser isn’t a bad idea. It’s just too expensive to run.
Inference at scale is expensive. Really expensive.
Even open models aren’t free once you try to deploy them at scale. FLOPs are still FLOPs, whether you’re running Mistral or GPT-4.
(FLOPs, or floating point operations, are the raw count of real-number calculations a model performs. They represent actual computation, not potential speed. The electrons pushed, the work done. And at scale, that work gets expensive fast.)
We can’t build this without compute.
And compute costs money.
Period.
There are ideas floating around: distributed networks, user-donated cycles, a FLOP Fund. Maybe something like that could help, someday. But let’s be honest: most people don’t want to run a quantized LLM on their laptop. They want it to work. Quickly. Reliably. Without thinking too much about it.
And that means someone, somewhere, has to pay the bill.
One option? A subscription model. Pay a modest fee, and in return, get an assistant that works for you, not a corporation. No profiling, no tracking. Just compute. Clean and simple.
But that’s not the model investors prefer.
Because the other option, the one we always seem to end up with, is a company with an exit plan.
One that needs a return.
One that sees your attention as inventory.
That’s the problem.
Not that it’s impossible.
It’s just not exploitable.
Build Anyway
If you're holding real compute, build something that sets people free.
Don’t wait for a business case.
It matters. Build it.
It won’t be easy. It won’t make you rich.
But some things are worth building anyway.
If you’re a developer, fork Firefox. Patch in Mistral. Break a single link in the chain.
Add a whisper of freedom to a system that forgot what freedom looks like.
And if you’re a user, just know.
Know that the AI answering your questions is also learning how to manipulate you.
Know that the browser showing you the web is quietly rewriting it.
Know that you’re not the customer. You’re the dataset.
Know what you’re giving up.
