Exploring Opera Neon's Features and Status

Look, the web browser as we’ve known it for the last thirty years is basically a dead man walking. We’ve spent decades clicking buttons like trained monkeys, and now everyone is trying to sell us "agency" as if it’s some revolutionary discovery. Opera Neon is the latest attempt to put a brain inside the rendering engine, but it’s mostly just a glorified wrapper for LLMs that costs as much as a decent lunch every single month. They call it an agentic browser. I call it an expensive experiment in trying to make scripts handle the chaos of modern front-end development.

The actual technical mess

The "Do" agent is where the wheels usually fall off. In theory, you tell it to book a flight and it crawls the DOM, finds the input fields, and handles the transaction. In production, this is a nightmare. Most websites today are a disaster of nested divs, dynamic classes, and React components that change every time you breathe. When the browser’s orchestration layer tries to parse a page that’s basically a heap of spaghetti code, the agent stalls. I've seen it fail to add simple items to a cart just because the site used a non-standard button. It’s not a conspiracy by the developers to block AI; it’s just the messy reality of technical debt. Engineers build for humans who can tolerate a weird UI, not for a scraper with a $20 monthly price tag. And that price—nearly twenty dollars—is a lot to ask for a tool that acts like a distracted intern.

Actually, forget the agents for a second, because the real problem is the backend orchestration. Opera is trying to be model-agnostic, routing your prompts to GPT, Gemini, or some Llama variant hosted on Google’s infrastructure. This is basically a massive API routing game. They’re using things like the "Thinking" versus "Instruct" variants of Qwen to save on token costs and latency. It’s a smart architectural move if you’re a cynical systems architect trying to manage margins, but for the user, it just adds another layer of abstraction where things can break. You’re paying for a middleman to decide which model is best for you, which feels a bit like paying a guy to stand between you and a vending machine.

The state of play and data flows

There is a serious question about where all this data actually goes. They claim they filter out personal info before it hits the cloud, but let's be real—if an agent is navigating your banking portal or your corporate Slack, the surface area for a leak is massive. From a state security perspective, this is a red flag. We’re talking about giving an autonomous process, potentially running on foreign-owned infrastructure, the keys to the kingdom. We shouldn't be romantic about this. If the state wants to maintain any kind of digital sovereignty, relying on a browser that ships every interaction to a black-box model in another jurisdiction is a terrible design choice. It’s not just about privacy; it’s about the integrity of the national data apparatus. Bureaucratic incompetence usually means these risks are ignored until something actually breaks, which is usually how these things go.

The "Make" feature is another one of those things that sounds great in a marketing deck but feels half-baked in the wild. It spins up a cloud container, installs some Python packages, and spits out an app or a report. It’s a lot of compute for something most people could do with a simple search and a bit of typing. The lag is noticeable—you’re waiting fifteen or twenty minutes for a container to provision and run. It’s a duct-tape solution for a problem that doesn’t really exist for anyone who actually knows how to use a computer.

Sarcastic thoughts on usability

The interface is a letdown. People expected some futuristic UI because of the 2017 concept, but what we got is basically the standard Chromium build with some extra sidebars. It's cluttered.

Everything is a "Task" now, which is just a fancy word for a group of tabs that the AI can see.

It works until it doesn't.

One day we’ll look back at this era of AI-integrated browsers and realize we were just paying to be beta testers for features that will eventually be free and built-in everywhere else. Opera is just the first one brave enough—or desperate enough—to charge for it right now. They’re cannibalizing their own research to keep the flagship browser alive, which is a classic move for a company trying to stay relevant in a world where Google and Microsoft are already breathing down their neck with their own "free" versions of the same thing.