Ask Paul: October 10 ⭐

Happy Friday! Let’s put yet another controversial week behind us and kick off the weekend a bit early.

? Windows 10, we hardly knew you. Now please leave

helix2301 asks:

What are your feelings on the EOL of Windows 10 next week and Office 2019? Windows 10 was a big breakthrough, I think, for Microsoft. It went from being the last OS to the more open platform in the span of about 10 years. Thanks to Microsoft changing CEOs.

I have mixed feelings about this, but I think that was probably true each time a major Windows version hit end of support. For people who follow Windows closely, these releases are rollercoasters of highs and lows. But they impact mainstream users, too, and the coming end of the Windows 10 era has garnered more attention in that space than I feel is warranted.

Whatever anyone’s feelings about Windows 10, one thing is objectively true: It is one of only a handful of Windows releases—the others being Windows XP and Windows 7—that were so widely used at the time of their exit from support that Microsoft was forced to adjust the schedule in some way. With Windows 10, we see the most extreme steps yet. In addition to supporting the release for businesses with an additional three years with paid support, Microsoft opened up additional paid support for consumers, too, albeit only for one year. And then things escalated. Anyone can get Windows 10 support for an additional year for free.

All year, we’ve speculated whether Microsoft would be “forced” to keep supporting Windows 10. I see both sides on this one. Windows 10 and 11 are close enough technically to be considered identical from a servicing perspective, so what’s the harm or cost in continuing to deliver security updates to Windows 10 users? On the other hand, the Windows 11 hardware requirements establish and newer and more secure baseline for supported PCs. And 10 years of support, 13 really, is a much longer support lifecycle than is offered by any mainstream personal computing platform (meaning the Mac, iPhone, iPad, and Android in all its forms). When you consider that other platforms with more users than Windows 10 are supported for less time, it’s unclear how this warrants so much outrage.

Looking at it as a user, but also as an amateur historian of this product line, Windows 10 feels like the product that the previous team would have made had they stuck around. (This was true of the shift from Vista and that team to the Windows 7 team as well.) They were already unwinding the most egregious of Windows 8’s sins and working to bring the PC and phone closer together. Windows 10 might have been called Windows 8.2 but for the horrible reputation that 8 earned.

But Windows 10 marked a nice return to desktop centricity, the correct decision given the nature of the user base. The problem, and this was no one’s fault, really, was that the merging of PC and phone, and related products like Surface Hub, Xbox, and HoloLens came too late. By the time Windows 10 arrived with its Windows Phone-inspired UI, Windows Phone was already dead. And that interface got stale fast. The WinUI 3 designs we see now in Windows 11 are much, much nicer and feel more modern.

At a platform level, Windows 10 suffered from having to continue forward with the mobile apps platform foisted on us by the Windows 8 team. And that continued to fail, with even its inevitable shift to a desktop focus in the Windows App SDK being an unmitigated disaster if my experience is any proof. (This thing is an absolute s$%t heap, in my opinion.) Not that it matters, in some way, given that app development shifted to the web and mobile platforms long ago anyway

Windows 11 has its controversies. But today, its once suspicious hardware requirements belatedly make sense, and the PCs made in the past several years are much more secure than their predecessors, especially Copilot+ PCs. It’s time to move on from older, less secure hardware platforms. The needs of the many, etc. etc.

We’re in a weird place. On the one hand, Windows doesn’t feel like a priority, and we are bombarded by new features on a monthly basis in a scattershot way. But on the other, Microsoft is transitioning to a more secure footing across the company, and Windows is getting what feels like incommensurate attention in that effort. So in addition to the hardware improvements in modern PCs, we’re actually getting meaningful security improvements in Windows at a platform level. I guess there’s always good and bad.

Anyway, I feel like it’s time to walk away from Windows 10. Microsoft has given consumers and businesses ample warning of the schedule, has extended support for both to some degree, and has made Windows 11 more attractive to those who care about security. Windows 10 isn’t really exiting support this month. It’s exiting support in three years. And that’s enough time.

✨ You can’t spell “paid service” without “AI”

wright_is asks:

With your view on “I won’t pay for AI”, how do you feel about the new Opera browser? You mentioned that reviewers get 90 days free, but would this be something you would pay for long term, or does you initial statement stand? Or is this something new, where you will have to reassess your position during the testing?

Aside from being an overt hypocrite, I think I can explain where I’m at on AI and why I think my stance hasn’t really changed without coming off like Obi Wan Kenobi and his contorted “from a certain point of view” justifications. Maybe.

If you think back to a bit over a year ago when I wrote I Will Not Pay for AI (Premium), my central issue was that Microsoft had created new subscriptions—Microsoft 365 Copilot for businesses and Copilot Pro for consumers—that one had to pay for in addition to their existing Microsoft 365 subscriptions. This felt like a money grab, and still does, because the features we had seen in those additional paid subscriptions were identical to the types of improvements Microsoft had been adding to Microsoft 365 “for free” as part of our existing subscriptions. This was the value and point of Microsoft 365: You subscribe, at least in part, because the versions of the apps we used with a subscription improved over time whereas the standalone apps and suites did not.

This shift to an additional per user, per month subscription fee is clearly enshittification because of the outsized cost vs. the value received. If Microsoft had simply made each subscription $5 more per month per user (or whatever), I don’t think I would have had much of an issue with that. But for consumers, especially, $20 more per month per user seemed insane, and still does. Microsoft 365 Family was $99 per year at the time and supported 6 users. So that’s like $8.25 per month for the full subscription, or less than $2 per month per user if all six slots are used by different people.

Things change. Microsoft 365 Copilot and Copilot Pro were and still are classic “copilot” or “besides app” interfaces in that they are tacked on, in a sense, and typically accessed in a sidebar next to a traditional and familiar app like Word or Excel. But these things are also evolving pretty rapidly and Copilot functionality is becoming more and better integrated with these familiar apps. To me, that doesn’t change the equation. Those features being more integrated is still the same thing Microsoft used to give us “for free.”

What’s changing now, however, is more profound and I think the way to frame this new shift is in terms of agentic AI, where these software agents are working on our behalf, as we always say, orchestrating ever more complex workflows using a growing number of backend services. This shift requires apps and services to be more programmatic, meaning that the AI agents can access individual features externally and combine those capabilities across multiple apps and services to get whatever tasks completed. And that is a different world.

It’s not there yet, not really. We will see half-steps where an AI-powered web browser literally moves the mouse cursor around, or whatever, and operates itself as if it were a person taking these actions, clicking on buttons, filling out forms, and so on. But the future of this behavior is hidden from view, it’s backend services accessing other backend services and coming back with results. Doing things like browsing for the best deal on some product and, given enough experience and trust on our part, buying that product at the best price. This is a leap of faith the current AI implementations do not warrant. But it will happen.

Regarding Opera Neon specifically, no, I will not pay for this. Not now. But we’ll see how this product and others evolve. It’s silly to say never, but where things are right now, no, I will not pay for AI. That could change. But the thing we don’t know is what this world looks like in the next year or so. These capabilities won’t be confined to Opera Neon, they will be everywhere. And in this hopefully competitive landscape of the future, we will have choices that include free and paid options. Perplexity Comet, which is very similar to Opera Neon, started off being limited to a $200 per month subscription, then it went to the $20 per month subscription, and now it’s free (but with limits, which is reasonable). This will be true of Neon at some point, too, I bet.

As I write this, I have three paid AI services that I get for free, all temporarily: Google AI Pro because I have a new Pixel (for one year), Perplexity Pro because I signed up via the Samsung Galaxy Store version of the app (for one year), and Opera Neon (for three months). And I have a sort of fourth version through Microsoft 365 Family, though I never use the features and never exceed the limits of the free capabilities when I use Designer to create images. I would not pay for any of these things if that was the option. And it will be, for most of that. (If I could hand over the Copilot capabilities I get in the Family subscription to my wife, I would. She does use Copilot and I don’t.)

The other thing I wrote that ties into this, I think, is When Everything is AI, Nothing is AI (Premium). That’s over two years old now (!), but it applies here. As these agentic AI capabilities roll out, they will go to paid services first. But they will come to free tiers. And while Big Tech can be trusted to keep squeezing the vice ever tighter by taking away more and more from the free tiers over time, competition might help lessen that. Because there will always be free alternatives. And they will get better in time, too.

Put simply, we can’t see the future clearly enough to say what we will or won’t do a few years down the road. And while I am open to paying for any service that delivers enough value, I strongly suspect I will never have to pay for agentic AI services. Perhaps there is a mirror of this scenario in my use of Typora instead of Word, or in Synology Drive instead of Google Drive or OneDrive. I did pay for both, yes, but only one time and not as a never-ending subscription fee.

We’ll see. But for now, no, I will not pay for AI, still.

⁉️ Intent

JustMe asks:

I have a follow-on MSA query. My question is about Microsoft’s stated reasoning for removing local account creation workarounds from the OOBE (not asking about reasons for/against MSAs – your article this week outlined those clearly enough). I keep going back to Microsoft’s own blog post.

In their own words from that post (about halfway down), Microsoft are removing local account workarounds from the OOBE because they are, “…potentially causing users to exit OOBE with a device that is not fully configured for use.”

I have to ask – what isnt being configured? In the interest of being fair to Microsoft, the only thing that immediately pops to mind is BItlocker (I am not including Microsoft’s screens that might try to sell you M365, XBox game pass, or other Microsoft services here.) Seriously, what are we missing?

I will try to answer this (see below). But first…

This is really bad timing for me, not your question per se, but rather this change because I am working on a 25H2 edition of the Windows 11 Field Guide that has a new format and structure, and I’ve been trying to make sure that it makes sense across different types of chapters/content. This requires ongoing refinements as I run into one-off issues in individual chapters. And though I would prefer to stick to new content for the initial release of this edition, I also need to look at topics, like installing and upgrading to Windows 11, that are classics in the sense that I revisit them every year.

Until this past week, nothing had changed with regard to the workarounds one could perform to bypass things like the Windows 11 hardware requirements and installing the system while offline (which is the underlying capability that allowed one to create a local account in the OOBE). This was comforting to me because it meant a lot less work. I have tested the known workarounds, and everything worked the same as before. Good.

But now they don’t. Or, they soon won’t. And that means I have a lot of work to do. I have done some of it. But not all of it, and that’s partly because this just happened: I can put Dev or Beta on one or more PCs, but Microsoft hasn’t shipped ISOs that include this new change, and so it’s difficult to fully test a clean install scenario. The best I could do is to install Dev and Beta and then reset the PC. That’s one of the things I’ve not done yet, but will. As soon as today, but probably tomorrow.

Here’s what I know right now. All the hardware requirement workarounds still work, and that’s handled by my latest Upgrade chapter or by using Rufus or whatever to create installation media. And that might be all I know for sure. Microsoft’s description of “removing known mechanisms” for using a local account during OOBE almost certainly apply to the “Shift + F10 followed by the oobe\bypassnro command line” workaround. That’s what I’ve used in the book because it’s so straightforward, but there are other ways. And that’s what I need to test. I’m literally working on the new Install chapter and was before Microsoft announced this change. So this chapter is going to be changing too, once I’m sure which workarounds might still work, if any.

The good news, as noted in my article this past week, is that you can still use a local account to sign into Windows 11. It will just take a few minutes longer to set that up. This is not the end of the world, and the complaining I see about this is bizarre to me.

Anyway, to your question. What do you miss from a PC configuration standpoint when you set up Windows 11 to use a local account compared to doing so with an MSA or Microsoft Work or School account?

It’s been a while since I’ve done that, and as noted, I will try to do so soon using the Dev or Beta channel build that kills some/all of the workarounds. But I wrote about this almost a year ago, mostly as an experiment to see whether it would help with any of the enshittification issues in Windows 11. This article doesn’t detail the OOBE steps you see and don’t see, but some of what’s missing, configuration wise, is tied to the PC being offline during the process while others are tied to the use of a local account. The most obvious and important is the one you mention, automatic full-disk encryption, which is transparent and probably not something most are even aware of.

I keep pretty much everything, so I went back to my archives from October 2024 to see if I could find screenshots of the full OOBE experience when signing in with a local account. And I literally have over 400 OOBE screenshots overall, thanks to me updating the book a year ago for 24H2. Oddly, I don’t have too many of this exact process. But I think that’s because it literally skips most of the OOBE.

So this is a rough list, it will be incomplete and maybe even incorrect in some ways. Again, I will try to figure this out more accurately soon. But …

• No full-disk encryption
• No PIN, unless you create a password for the local account
• No Windows Hello, per above
• No automatic account recovery
• No data recovery with versioning through folder backup because there’s no OneDrive configuration
• No ability to restore from a previous backup

There are other things, of course, that are either less important/useful or are overt Microsoft ads for things like Game Pass and Microsoft 365. But I feel like those are what Microsoft was referring to. And it’s mostly legit.

One final point tied to this.

I knew going into this topic earlier this week that some would be violently opposed to the change. I knew that some would describe this, inaccurately, as enshittification. I knew that me explaining why this is not enshittification would not satisfy those people. And I was thus unsurprised by the feedback from that group, even the personal attacks. One person literally accused me of having a “conflict of interest,” which is objectively ridiculous given my history. But also hilarious.

(As I wrote in A Windows 11 Enshittification Checklist (Premium) about 18 months ago, the forced MSA sign-in is “Not an issue. Mainstream Windows 11 users should sign in with a Microsoft account. Those technical enough to not want to do so can figure out how to workaround it.” This is still true.)

The problem I have with all this is tied to a problem I see all over the place, which is that the world is not black and white. There is nuance. And in the sense that we should see nuance in a Big Tech abuser like Google, which has the best search engine and is also destroying competition and harming its users, we should look at this similarly. That is, some of what Microsoft is doing here is self-serving, for sure. But this is a two-way street. And the real world benefits of signing into Windows 11 with an MSA vastly outweigh the downsides. You can easily ignore or uninstall all the MSA tie-ins you don’t like if that’s where your head is at. But the justifications for signing in with a local account are feeble, they’re like complaining about wearing a motorcycle helmet because you don’t want to mess your hair. The rationale for signing in with an MSA isn’t just solid, it’s a no-brainer. And that’s true no matter how cynical one is about Microsoft.

? There can be only one … web browser

jrzoomer asks:

Paul where do you currently stand on browsers in late 2025? Are you on Chrome or Edge or Brave these days mostly? I know Arc was your go for much of last year but not sure if it still is? How about Opera/Opera Air, Vivaldi, Sidekick, DuckDuckGo, or Perplexity’s Comet? And last, is Firefox dead to you? Oh and there’s also Zen, which is based on Firefox?

All things being equal, I use and recommend Brave. Full stop.

But I also have to use different browsers from a sort of professional perspective, so I move between browsers all the time. I’ve used Edge a lot this year mostly so I can experience and then document what’s going on there with Copilot and whatever else. But I’ve also used all the browsers you’ve mentioned, including Firefox and Zen, neither of which I can recommend or want to use because of performance issues and the silliness of trying to innovate/compete with a rendering engine that’s the wrong place for that. And others, like Dia, which I do like quite a bit and can’t use regularly because it’s Mac-only right now. (And Safari on Mac, iPhone, and iPad.)

Honestly, you would probably be fine with almost any browser assuming you protect yourself with the right tracker and ad blockers and have enough RAM. I find that Edge and Chrome are particularly resource-heavy, and that’s a big problem on PCs with just 16 GB of RAM. Brave performs better than Chrome and Edge, and obviously, on the low-end HP OmniBook 5 (with the lowest-end Snapdragon X processor and 16 GB of RAM) I’m using to write this. In fact, those browsers are almost unusable with a lot of open tabs.

But other than that, and any worries one might have about Big Tech stalking your usage, they all work similarly. This will change as AI advances and these browsers shift into more agentic apps. But for now, nothing has changed.

? I, Robot

ianceicys asks:

Paul, wonder if you’ve seen the Figure 03 video on their new robots? The trailer seems super intriguing, even if this is 10-15 years out think of the implications. A trailer is not real, and so maybe this is all a mirage? Would love to hear your thoughts on what if this is only 5-10 years until we have robots in our home.

I’m vaguely disturbed by the notion that we will ever use human-like robots in homes. This feels too much like slavery on one level. Plus, we already anthropomorphize pets, other living things, and even inanimate objects. So we don’t need robots to look like humans, and any number of other form factors may be much more efficient. It’s also weird to me how slowly this thing walks around. I could imagine waiting at a table for it to slowly creep up with a tray of food or whatever and losing my mind because of how long it’s taking.

But it all comes back to the slavery thing: Every task that this robot performs in that video is servile in nature. And it’s the human shape that makes this uncomfortable, like the future will be us in a Downton Abbey existence commanding those lesser human(oid)s who live (recharge) under the stairs.

There’s no doubt that we’ll have human-shaped robots. But existing robot installs, like those in the automotive industry, aren’t human-shaped, they’re custom for the task. And that makes more sense to me than spending years developing something that looks, moves, and interacts with objects like humans do. It’s not like we’re the ideal form factor. I can imagine multiple devices, each optimized for specific tasks, and many much smaller than a humanoid robot. All with whatever shape makes sense.

What about Amazon Astro? It also seems like it could also be 50 years away. What I find super surprising is that Microsoft isn’t in this ecosystem even with Microsoft Robotics Developer Studio was discontinued. Would love to hear your thoughts.

This thing is ridiculous in its current state, but also kind of cute. And Amazon is a double-edged sword in that I don’t trust it to make anything high-quality, but I also appreciate that it’s trying, and doing so this early. I also like that Astro is more of a smart home device, like a robot vacuum, than a humanoid shape. I would enjoy seeing it try to navigate stairs, but at least one robot vacuum model actually does that now, so anything is possible.

We’re early days on robots in homes, obviously. Smart home technologies are mostly unnecessary today, so there’s that whole mess to navigate and lots of hurdles to get by. But this thing, oddly, feels closer to how robots in homes will look and work. Maybe it’s just because this is more viable in the short term. But I do think the non-humanoid route is correct or at least better.

Regarding Microsoft, it occurs that we’ll suffer through generations of home robots from companies like Apple and Google (and Amazon) that have strong consumer bases, with all the expected lock-in and incompatibility nonsense too. So that will be fun.