I spent the weekend using Firefox Smart Window as my default work tab setup, with a regular Private Window sitting next to it for anything I did not want an assistant to see. The short version is that this is the first AI browser feature I have not disabled after a day. The longer version is that some things work the way I expected, one thing I thought would be marginal turned out to be the reason I kept going back, and the privacy story is real but conditional. I want to walk through what actually changed in my workflow so you can decide whether the waitlist is worth your email.
What Smart Window is, in one paragraph
Mozilla has quietly added a third workspace option to Firefox, distinct from the Classic and Private windows you already have. It is not a separate browser, and it does not replace anything. Think of it as an extra room in the same house, with an AI assistant wired into the walls, ready to work with whatever you bring in. Mozilla is rolling it out slowly in the United States, Canada, and France, with a waitlist open for everyone else. I tested this in a recent beta build on a Windows machine.
The reason the distinction matters is that anything you open in this window is fair game for the assistant to read, summarize, and remember across sessions. Anything you open in a Private Window is not. That is the rule I set for myself on day one, and it is the rule that has not changed.
What got faster
The honest report is mixed, and I would not trust a review that only listed wins.
The headline win is multi-tab research. If you ever price-shop a laptop by opening four retailer tabs and trying to keep the specs in your head, you already know the slow part. Smart Window’s assistant compared four tabs for me on a real shopping run and I trusted the result because the source tabs were visible next to the answer. The same worked for a flight search where I had three dates open in three tabs and wanted to know which combination was cheapest across them. I clicked through and confirmed. That workflow is genuinely better than copy-pasting numbers into a spreadsheet.
What I did not expect to be the standout was tab housekeeping. I do not lose much time to a single duplicate tab, but I lose noticeable time to the pattern of opening six tabs for one task, leaving them open, opening six more for the next task, and then hunting through eighteen tabs for the one I actually need. The assistant grouped related tabs for me without my asking, which sounds small and is not. That is the feature I kept using after the weekend ended.
What got slower, and where I bounced back to a regular tab
Anything single-tab. If you are reading one article and just want to keep reading, the assistant is invisible and that is correct. The cost showed up only when I tried to ask a quick question about a page and the answer took a beat to come back. The latency is fine for research, annoying for what does this one paragraph mean. I stopped using Smart Window for the latter and bounced back to a regular Classic window.
The other slowdown was opening a sensitive account by accident. Smart Window is not Private Window. Anything you open in it is in scope for the assistant. I caught myself about to open a banking tab in the wrong window once, and that was the moment I switched to a strict rule: Smart Window for research, Private Window for accounts. Not a criticism of the feature, just the kind of rule you need to set for yourself before you start using it.
The web retrieval piece, and why I trusted some answers but not all
The part most reviewers skip is the part that decides whether this is a real product or a polished demo. Smart Window’s live web search runs through Exa, a search provider built specifically for AI workloads, and every answer that uses live data shows the source links inline. That is what let me click through and confirm the laptop and flight comparisons above. If the answer is grounded, you can see it.
The trade-off is that Exa is not Google, and the index is smaller. For niche local business searches I had to fall back to a regular search tab. That is a real limitation. The good news is that the inline sources make the fallback obvious. You see the answer, you see where it came from, and you know when the answer is thin. That is a better failure mode than most AI browser features.
The privacy posture, and why it is conditional
Mozilla has been explicit that the assistant does not use your browsing data to train third-party models and that the Exa queries are not tied back to your Firefox account. That is a meaningful distinction from competitors that train on user data by default. Whether you believe Mozilla’s policy is up to you, but the policy itself is public and written in language a normal reader can parse.
The catch is the word currently. This is a beta. Features evolve, the privacy page can change between now and the wider rollout, and the version you join today is not guaranteed to be the version that ships to everyone next year. I would extend trust to the policy as it is written today, and I would re-read the privacy page every few months for as long as I keep using the feature. If the policy changes in a way I do not like, I will stop using Smart Window. That is a tradeoff I am willing to make for the productivity gain, and you should make it deliberately rather than by accident.
What to try it for, after three days
Here is the test plan I would suggest if you are on the waitlist and the email arrives tomorrow.
- Pick one research task you have been putting off. A laptop comparison, a trip plan, a vendor decision. The kind of work that has you opening ten tabs and closing nine of them.
- Use Smart Window for that task and only that task. Keep your other browsing in your normal windows. This isolates the experience so you can tell what Smart Window actually contributed.
- Verify one of the assistant’s answers by clicking through the source. This is the part most people skip, and it is the only way to know whether the feature is actually grounded in real data.
- Decide whether the cross-tab comparison saved you more time than the duplicate-tab grouping. For me the comparison was the headline feature. The grouping was the one I kept using.
That last test is the honest one. The assistant’s headline features get the press. The one that changes your daily workflow is usually the unglamorous one.
Trade-offs
The honest trade-off is between a real productivity gain and a privacy posture you have to actively maintain. Smart Window gave me back noticeable time on the kind of research work I already do, and the privacy story is better than the alternatives I have tried. The cost is that I now have to think about which window I am in before I open anything sensitive, and I have to remember that the privacy story I am trusting today is the privacy story as it ships in beta, not as it will exist in a year.
The other trade-off is between Smart Window and just opening more tabs. Most of what the assistant does, you can do yourself with more discipline. If you do not want another AI surface in your browser, you do not need this. If you want one and you already trust Mozilla more than the alternatives, this is the version of that idea I would actually use.
What I would tell past me
Three things, if I could go back to the morning I installed it.
- Set a rule for which window handles what before you start using it. Smart Window for research, Private Window for accounts. The rule is more important than the feature.
- Verify one answer a day by clicking the source link. It is the only way to know whether the assistant is grounded or just confident.
- Do not trust the privacy story past the current beta. Read the privacy page again in three months and again in six. The feature is allowed to evolve.
If you do those three, you will get the benefit of the assistant without losing track of what it is doing with your data. That is the version of this product that earns a permanent spot in your workflow.