Ask Paul: February 28 (Premium)

Happy Friday! Here’s a quick update on this week’s site update, in addition to another great set of reader questions.

? Quick update on the site update

I’d been meaning to update the site header, at the very least, for a while. Like a long while, maybe as long as a year. But owning this business has been a painful reminder and/or example of that adage about system admins and IT pros always having to be reactive while wanting to be proactive. That is, life always gets in the way. Not my personal life, usually, but just everyday work things that I was sheltered from for most of my career. Things that need attention, are usually a surprise, and an unwelcome one, and are almost daily occurrences. All I want to do is write. But there’s a business and it needs attention, and I hate it.

Anyway, this is just mostly an excuse, not a reason. As I writer, I’ve settled into a long-term rhythm of sorts, and I don’t like the wrenches that get in the way of that. This is tied to the AI conversation below, as well. I just do the things I do, the way I do them. But I can’t many days, and it gets frustrating. I’ll get there. Or I won’t. I don’t know.

Getting finally to the point, I had wanted to update the header for a long time. I’ve liked the Thurrott.com home page since I saw the first design, probably a bit over three years ago, but it was also sort of done without a lot of my involvement. And there are things I would do differently. And can, now that it’s my site. I just haven’t. And now I finally am. This will occur in phases, but starting with the home page made sense to me. And in looking at the site header, and thinking about what I wanted to promote up-top, it occurred that there was more to do, perhaps.

Some of this is perhaps personal preference, but basically tightening up the white space, which has its place but was I think a bit too much. The home page is constructed of blocks, and so this was basically about reducing the height of things, both inside and between those blocks. Make the header less tall, for example, while also adding some obvious links to related things in a colorful, button/icon-like way.

When I finally got started on this, Robert, the genius I am so lucky to know and work with on the backend site tech, wanted me to be as specific as possible. And so I sent him an itemized list with mock-up graphics for him and his team to work off of. Here’s a shortened version of that.

Site header

Bigger “T” logo and Thurrott logo, Thinner (less tall) and wider search bar, Icons for Premium, Thurrott YouTube channel, the Windows Intelligence newsletter, and Eternal Spring. Remove the “Contact” and “About Thurrott.com” text links. Remove name next to the user avatar image. (I joked that the person signing in to the site knows who they are, and this is just taking up space.)

My first pass was just to use Office icons to show roughly the layout I was considering.

And then this.

Among the issues there, we never made a real Thurrott Premium icon or whatever. We had used different diamond images here and there, so my first shot at that icon was to use a diamond shape. But then everyone involved complained the icons were too different and didn’t line up visually well. So we changed that. Well, I did, mostly.

Content blocks (all)

Heading should be bigger (larger font), and I want a way to control which blocks appear on the home page.

Forums block

Each entry should be thinner with no icon. I had an idea about the layout.

Footer

Smaller, less tall, with fewer elements since almost no one even looks at it in the first place. The elements we keep will include the Windows Intelligence sign up box, Thurrott logo, Copyright. About, Contact help and support, Email Paul, and RSS.

And … yeah.

I spent about an hour with Robert going over this about 10 days ago, and he made some changes immediately. Including the temporary new forum block, which wasn’t quite what I wanted, but we ran out of time. But then yesterday, a teammate of Robert’s started implementing the remainder, and I realized that the icons I had come up for the header weren’t right. So I spent most of yesterday morning just working on that, going back and forth with Laurent, the newsletter folk, and Robert and his team. And so we arrived at the more consistent design for those buttons that’s up there now. Perfect? No. But better, and overdue, and I’m glad to get it to this point.

There’s a lot to do, I know. I will get started on the section pages and article pages soon, to make it all consistent with the home page. There are a few internal things, like a management interface for which blocks appear where, to sort out. And then whatever else.

But I just want to write. So let me get started on that for the day.

?️ Legacy

helix2301 asks:

Where do you think Windows and server enterprise will land going forward? I know those are legacy businesses for Microsoft but they make A LOT of money. There are still LOTS of business customers on Windows and Windows Server etc. We all know as long as Microsoft is making money they will continue with it but let’s be real they may push servers and stuff to the cloud but Windows itself has billions of users Apple 100 million users on MacOS. It’s like that legacy business that Microsoft can’t live without really what operating system would everyone use Linux or chromeOS. Windows is kind that gateway to sell you office and everything else Microsoft not really legacy its gateway product.

In some ways, the situation here mirrors things we’ve seen in the past. For example, while Apple was pushing the Mac almost exclusively from 1984 through the end of the 1980s, it was the Apple II line that the company rarely mentioned that was bringing in all the profits and revenues, and keeping the company afloat. Microsoft has, of course, worked similarly over the years, pushing Windows when MS-DOS was king, etc. Technology is always forward-leaning, but when you combine that with the growth needs of Wall Street, you get into a sort of wag the dog situation. And in today’s world, corporate reporting is so vague, it’s very easy for Microsoft and others to hide where the money is coming from. We’re at the point now where quarterly reports are as much about the future as they are about the time period in question.

(Related side-trip. I didn’t write about either of these when I covered HP’s and Dell’s latest financial reports last night, but both were an example of the above. In HP’s case, the real story this quarter was that the company is on a cost-cutting crusade over several quarters and will benefit from that over time. And they are moving most of their manufacturing for North American PCs outside of China. This is something Lenovo won’t do, for example, and HP, unlike Lenovo and Dell is just PCs. Those companies both do servers as well. For Dell, the story was that they’re sitting on what I think was $8 billion in unrealized datacenter server revenues because they just can’t get the parts. So their future upside is great, and Wall Street loves that. For me, it’s all about PCs, but there’s almost nothing to go on there that’s good or interesting, at Dell.)

So there’s that. But starting with Server, the shift from on-prem to cloud is also a mostly one-way street. In a sense, that does parallel the MS-DOS example from above, when you think about it. For a while, there will be both on-prem and cloud-based solutions, and Microsoft is strongly positioned for that hybrid world. But in time most servers will be cloud-based, not on-prem. I know servers will never really “go away,” just like COBOL never really went away. But they will go away for the most part. And for on-prem solutions, whether it’s individuals, small businesses, or whatever sized enterprise, that will shift to either cloud-based entirely or with some mix of more specialized on-prem hardware. Something like a NAS, actually. I went through an unreal series of server-type products at home, including for a while, literal rack servers, but simpler wins in the end.

If you look at what Microsoft did with Windows Server 2025, there’s an interesting mix of more of the same–a lessening of standalone on-prem workloads, lots of hybrid capabilities–and a bit of a readjustment based on real-world needs and customer feedback. We’re not going back to on-prem servers as a thing. But Microsoft might have pushed too much and too fast to get companies off these things when either they can’t (for regulatory reasons) or won’t. Or are doing so on their own, slower schedules. So WS2005 is a response to that reality.

When I examine Microsoft’s earnings each month, I tend to stick with those products/technologies I care about the most (client, mostly, but AI too recently). But the most recent analysis piece I wrote (I hate that term, but it just makes sense for titling purposes) took on a life of its own, and I ended up focusing almost exclusively on AI. But looking at the content for that quarter on Microsoft’s investor site again, I’m again reminded of this hiding of details and inability to get a real picture of what did what. In one sense, you can argue that roughly 80 percent of Microsoft’s revenues now from businesses it would describe as cloud-based. But it’s more nuanced than that. (And Microsoft’s so-called Microsoft Cloud non-business accounted for about 50 percent of total revenues, though we’ll never know products are in there.) So how much did on-prem contribute?

There’s no way to know for sure. Server products and clouds services are commingled and not called out individually. Azure and other cloud services revenues never get a hard number, so we don’t know what the percentage of all Intelligent Cloud revenues that makes up. (Microsoft claims that the 19 percent overall revenue growth in that business was “driven by Azure,” but that’s growth, not actual revenues.) We know that smaller businesses tend to grow faster than big businesses, so does that weigh in when Azure had 33 percent growth and Server products had just 3 percent growth?

Maybe not. Nvidia is the second-biggest company in the world and its biggest product by far, AI GPUs for cloud datacenters, is also experiencing dramatic growth. Meanwhile, its legacy businesses are not. So it’s possible Microsoft is seeing something like that.

Possible. But unlikely, because it would explicitly tout that if it could. Round and round we go.

On the client side, it’s likely we’ve landed at what the PC industry is, from a unit sales point, and will remain for whatever number of years. IDC just changed its forecast for PC sales in 2025 downward a bit because of the ill-conceived new U.S. tariffs, but we can call it about 275 million units, give or take 10 million. Roughly. We don’t know what percentage of More Personal Computing revenues that Windows contributes–plus the commercial licenses that are now under Microsoft 365 in Productivity and Business Processes. But we can guess it will be stable for a while. Stable is good when we’re talking 10s of billions of dollars each quarter, and we are. But it’s bad for Wall Street, which wants growth. So we don’t hear about it.

Windows is impacted by two realities. The most obvious is that it’s just not a corporate priority. Microsoft wants consumers and businesses to consume subscription services (Microsoft 365, Xbox Game Pass, whatever) and it doesn’t care where they do so. This is a big part of the Satya Nadella strategy, to just meet customers where they are. And for Windows, this can work because most of that usage, and most of that revenue, is coming from businesses. Which are slow moving and will stick with what they know. So they can ignore Windows at the senior leadership team level for the most part. That business just keeps going. But this is why the consumer stuff is so terrible. It’s all B- and C-teamers internally, and who cares when the customer base is pretty small? Let them buy a Mac. As long as they use Microsoft 365.

The second impact is the broader, now established trends in personal computing where we all spend far more time with far more personal devices that in most cases fit in our pockets. PCs are relegated to mostly work-related functions that require a big screen and a real keyboard. And that impacts how individuals view those devices and Windows. It’s not positive. Aside from gamers, who probably want to get by Windows as quickly as possible, there’s not much there that matters from an audience size. And Microsoft seems to be actively plotting a shift to some Xbox/Windows hybrid for those devices anyway.

In short, both platforms are surviving mostly on inertia at this point. The PC will survive mostly because businesses are change averse. And Microsoft will continue mostly ignoring Windows at a high level because it’s not strategic. The Mac, for all its advances, is still just 10 percent of the market, and Chrome OS has not inched upward in a meaningful way in years. And that’s all you need to know about this market. It’s just inertia, with little in the way of excitement, even with those platforms that are perceived to be newer, or better, or whatever.

? AI is real, unlike Sasquatch and UFOs

spacecamel asks:

(Sorry for the rewrite of my question to sound less weird. Communication is hard.)

Not at all. In the communication is hard department, I saw that in my comments moderation interface and answered it before realizing it was part of Ask Paul. So I removed my response so I could handle it in more detail here.

Would you consider using AI to write some of the pure news articles so allow you more time to work on analysis? I know you are opposed to using it especially as a crutch but the tech is there and I think it does a pretty good job for a type of article such as your quarterly results articles.

I guess I need to be careful not to get into a “never say never” thing. By this point, we should all agree that AI is changing and improving so rapidly, that sticking to previous assumptions or even evidence-based opinions can quickly be out-of-date.

Today, I have no plans to ever use AI to write. That is, I previously described, in an earlier Ask Paul, how the summarize functionality that we see in things like Copilot, ChatGPT, and other AI chat interfaces is very much duplicating the work that bloggers have done for the past few decades. And that this functionality can and will replace those jobs. (As it should do immediately at Neowin, which has the worst writers on the Internet.) But I also resist this shift personally, not explicitly or in any denial of reality sense. I just do what I do. And time goes by.

The other day, I wrote about some of my hesitant first steps into using AI for work. None of the examples I gave involve writing text for me, they instead relate to ways in which I am trying to use AI to save time on tasks related to my daily work. (Which I had also written about recently.) This is an example of it being easier to give advice than to take it. I am paying attention to AI, of course, but it’s difficult to stay on top of something that improves almost literally every day. And so I read, and watch videos, and experiment.

And I’ve gone to my wife, who is also a writer, a few times and said, hey, this is kind of geeky or technical or whatever, but you should watch this video, or read this article, or whatever. And when I was prepping for a Hands-On Windows video about some AI features in Notepad, of all things, I showed her some of the advances coming to Copilot+ PCs soon because they are impressive and will be things we all use in the coming year or whatever. She was inspired by this to try AI for her own work, as I wrote about a few weeks back. And her real-world use of AI inspired me, in turn, to try and do the same. Hence, the articles noted above, which is me trying to walk the talk, if you will.

I go into this hesitantly. Again, not out of some explicit strategy. But because I’m a writer, and that’s what I do, and saving time writing isn’t my goal. My goal is to spend time writing. And less time doing other things I care about less. But that doesn’t mean AI can’t help me write better–which it does already, if you consider spelling and grammar checking services (I use LanguageTool now, but Grammarly is more common and well-known) to be some form of AI use. They are, of course, but perhaps the distinction here is between “normal” (traditional?) AI and generative AI. I use generative AI to make website images, for example, in part because that is not my core job or something I care to spend time on. But I don’t use generative AI to write. Because that is my core job.

One of my goals for this year is to use my work archives as the grounding for a custom GPT, or app, or whatever, that I can prompt to more easily found things I’ve written over the past 30 years. This will likely take the form of a Retrieval-Augmented Generation (RAG) solution in which an established AI like ChatGPT is grounded, rather than trained, on my data and is used as the sole source for the answers it gives me. But perhaps it will be as simple as semantic search in Windows 11 working off that data in OneDrive. We’ll see. But whatever form it takes, that will save me time, for sure. I get lost in my own archives all the time. (And will likely do so again this weekend, when I hopefully write a Skype obituary based on decades of my previous content about that product.)

AI is moving so fast. So much so that I feel bad for people who continue to deny its usefulness or, worse, actively take steps to prevent it from appearing on the PCs and devices they use; even worse still, they do so for their parents or others who trust them with technical matters. This is worse than a generational divide, it’s like claiming the earth is flat when that claim is instantly debunkable. It’s gone from an opinion to a delusion. And every time someone points out whatever flawed Apple Intelligence headline or whatever other AI mistake, I just think, man. You are ignoring the millions of successes that these things accomplish every single day to make some out-of-date point. The exceptions make the news. You never see the headline, “Man dies by heart attack despite wearing an Apple Watch” even though that happens thousands of times more frequently than the success stories you always see. (And not to be a dick about it, but look at the adults in this picture. You know what else would save your life? Eating less and exercising more.)

I’m getting off track.

Regarding your remark about a crutch, and tied to my unfortunate comment above about people who perhaps could be doing more for their health than over-spending on an Apple Watch, I think it’s important to put this is in perspective. I’ve struggled with weight my entire adult life, so I’m a hypocrite. But I’ve also always said that if someone would just invent a pill I could take every day to stay thin, I would take it every day for the rest of my life. And now we basically have that, through Ozempic and whatever else. And there are people calling those who take such things cheaters. This is unfair. If I know anything about health and weight loss, and I really do know one thing that is objectively true, it’s that we’re all different. And that means that you can read a story about how some guy lost a lot of weight and try it for yourself, and it won’t work. Not ever, but most of the time. There is no one way. Many of us are biologically or genetically predisposed to gain weight, and those that are not shouldn’t be pointing fingers. You don’t understand the struggle. Some people smoke their entire lives and never get cancer. As with AI, the exception doesn’t prove the rule.

Likewise, using AI is not a crutch and it is not cheating. I happen to be a writer, but most people aren’t, and I know fully grown and educated adults who write horribly. We have tools that can help and it is stupid, maybe even irresponsible, to not do so. AI is just the latest in a long list of advances that had their naysayers, from the fountain pen to the train to the airplane. And in each case, there were old timers with their tales of “back in the day,” or “when I was growing up” and … yeah, whatever. Life improves. Technology advances. I was born before personal computers. My kids were born before smartphones. Their kids, if they have them, will grow up in a world in which AI is a major part of the fabric. Comparing that to the black and white TV I watched as a child in the late 1960s isn’t just a generational rift, it’s two different planets.

But I struggle. I will continue using AI as I can. I will expand my use of AI, no doubt. I don’t have plans to let AI write for me. But if and when AI can write like I write, or I can train AI to sound like me and read something I wrote, or generative podcasts or videos automatically, maybe this becomes unavoidable, even desirable. Maybe not using it that way is stupid or irresponsible. Maybe that day is already here. I don’t know. All I can do is be open to it, keep experimenting, and see where AI can make the biggest impact in my life, professionally, but also personally. And just know that that usage will only expand, and dramatically.

By the way, I like the new icons that you have added to the site.

Thanks. I was going to write a bit about what led to that, but I pulled it out and put it up top as it got too long.

?️ Local vs. cloud

digiguy asks:

I see a lot of people online testing local LLMs, mainly Llama, on powerful GPUs or high-end Macs (because to make them viable you need a lot of VRAM / unified memory). Have you tested some local LLMs or do you see yourself doing it at some point? I suppose it does not make financial sense to spend a couple of thousand dollars on such hardware if you are not some AI developer, but if you have or buy the hardware for other reasons (gaming, etc) could it make sense to run local LLMs instead of paying for AI? (assuming that these can do better that what’s offered for free…)

I have, but not as much as I thought I would.

There are a few sides to this. There’s the developer angle, where you can write code that targets these capabilities. There is running what I will call a Small Language Model (SLM) locally, as opposed to an LLM in the cloud, for chat, generative AI, and so. And then there are apps that take advantage of these capabilities, which today are largely creative apps like video editors and so on.

On the developer end, I spent much of the past year waiting for the Windows Copilot Runtime to appear, as it targets on-device AI, specifically in Copilot+ PCs. But that didn’t appear until very recently, which lead to a quick write-up. But this was a good reminder that these things need to evolve into an orchestrator-based system in which the PC (or whatever device) intelligently pushes AI workloads to whatever resources are available, a local NPU, a local GPU, a local CPU where that makes sense, or the cloud, or whatever combination. Specifically targeting only PCs that have a certain class of NPU isn’t just limiting, it’s short-sighted. So that’s not a non-starter, per se, but it feels temporary.

The second one is what I think you’re really asking about, which is that you download some SLMs to your PC and interact with them using a chat interface, usually, though apps (see below) can do so as well. And you do this in lieu of using cloud-based AI for any number of reasons (cost, privacy, data sovereignty, etc.), all of which make some sense. And yeah, I have experimented with this. I do have a Copilot+ PC, but most of these local AIs target GPUs right now, and this is another area where these capabilities need to improve. But this will become more mainstream, and will improve generally, so we’re not far from the point where querying a local AI, however you do it, can make sense, even on mainstream hardware.

The third side will become the most common over time, I think. If you look at Copilot today, or any AI chat interface, it’s not difficult to imagine a UI switch where local AI is a choice or can be used in a hybrid way. This will happen as AI PCs become more widespread, and as Microsoft makes Copilot+ capabilities (for the PC, specifically) work with GPUs and CPUs over time. But also just apps, bundled with Windows and from third parties, that have AI features that work locally when that hardware is available and are cloud-based otherwise. Right now, Windows is kind of weird in that regard. Paint, for example, has several AI features, some of which work locally (and require specific hardware, for now) and some of which use the cloud. They’re not always clearly identified, but in time they won’t need to be.

All these are things I need to spend more time on. This isn’t a concern for most, I guess. If you use an app like Premiere Pro and it gets better when you buy a Copilot+ PC, that’s great. But that’s just a feature or the type of thing one would expect from any upgrade. But I need to be a little more forward leaning than most, and tied to all the commentary above, this world is moving so quickly, it’s hard to keep up. It’s hard to know where the best bang for the buck is (where the buck could be money, literally, or just functionality or time) because it evolves almost daily.

But if you want to experiment with local AIs, it’s pretty much doable on almost any reasonably modern PC or Mac. Whether that leads to something useful day-to-day or not is unclear. But like all AI, it gets better over time. Having a local AI that knows all about C# and WPF, for example, would be really useful to me right now. Perhaps there are more mainstream applications for this too.

? How far is too far?

jgraebner asks:

I’ve been following with interest your articles on improving the resiliency of your online accounts and have come to realize that my first step to doing the same should be to move away from using a gmail address as my primary email in favor of a custom domain.

I hate that I needed a wake-up call to do this. But this is perhaps human nature, and I’m a natural procrastinator, especially when it comes to something difficult, uncomfortable, or time-consuming. And this is all three.

But the central question here is about Big Tech. It’s inevitable, but we have choices about what role these companies can play in our lives. And so this is about how and where we will use their services. And how and where we will not. And in my own goofy pragmatic way, I am trying to find the middle of the road balance that makes sense for me. I am trying to not go off on some extreme tangent here.

For me, that middle ground is to acknowledge that Big Tech products and services are too useful and pervasive to ignore while trying to minimize my exposure to them too. Those seem like contrary aims, but my YouTube experience is instructive. I could have lost that content. I will now do what I can to make sure that never happens. But I will also continue using YouTube.

To your point, I’ve been wrestling with what I’ll call online identities for many years. I intended to logically consolidate my many online accounts as part of that digital decluttering push a few years ago, but ran into all kids of issues. And then Google took away cheap additional storage for Workspace customers, forcing one change. But there are so many issues there.

There is nothing more central to our online identities these days than the online account(s), and thus email addresses, that we use. (Actually, a phone number is curiously important, still, but at least one can quickly bring up a new phone using an eSIM to reestablish that connection should a device be lost or stolen.) And choosing which address(es) to use, and with which services, is a key to that. My thurrott.com domain is at Google for what I’ll call legacy reasons. And there is some lock-in there, both good (it’s a good service, really) and bad (it could be taken away at any time and for no good reason at all). But at least I own the domain and if Google goes berserk some day, I can get the email address back. Not the content, in that instance. But at least the address. It’s mine.

So that’s one level of protection I have that someone with a .gmail.com (or .outlook.com, or whatever) address does not have. What you’re talking about is the next step, and it’s a big one. To get a custom domain, host it somewhere, and move away from using a .gmail.com account as your primary online identity. This may be difficult, if my experiences are any guide (or I may just have personal limitations that made my experiences worse, I should acknowledge that). But it’s something that will require thought and planning.

It’s a good goal. There is probably some version of me using Google Workspace but hosting my email elsewhere. Or … me not using Google Workspace, which costs money every month and … gives me some kind of service. This is all very individual, I guess. I kind of wish that this email address (my primary identity, paul at thurrott.com) was not tied to my work so explicitly and could just be mine, personally, again. Some day. But that’s what you’re doing. So there are services to consider, and costs. There are no obvious choices.

You’ve mentioned in the article that you have been using a Google Workspace account for that, but do you have some ideas of what options you would consider if you were starting from scratch today?

Google Workspace or M365 Business each seem like obvious choices, but I’m a little unsure of the potential downsides of using either one as an individual rather than a business. Proton Mail was one of the first alternatives to come to mind, but I’m not really happy with the lock in to using their apps instead of something like Outlook. More dedicated email services like Zoho Mail or Fastmail also could be possibilities.

I haven’t thought about it too much, but, in my case, I probably would put Proton Mail at the top of the list. I likewise haven’t considered what that would look like from an app perspective (web app is fine on desktop, in my case, but would need to figure out something on phones, etc.). If you literally can’t configure a custom domain in Proton using Outlook or whatever apps, you could simply forward things through a Gmail or Outlook.com and use those apps. But keep all the content back in Proton. Or something.

The problem here is that Big Tech hits on that affordability axis as well as the usability/functionality axis, so this is a big step for most. I’m not there yet. But I’m going to add this to the Online Accounts to-do list as something to pick up when I ever get through the items that are on there. It’s definitely something to consider, but it’s one I’ve just not spent enough time on.

Regarding Proton, I use their password manager everywhere and it’s wonderful. And I use Proton VPN here in Mexico when I need to. But Mail and Calendar are obviously core to that family of offerings, and there are day-to-day reasons to consider them, including encryption, anti-tracking, and the other privacy functionality. They’re open source, too, which you’d think would lead to solid support from email services and apps. But my experience with each, so far, has been cursory. Just thinking about changing from Workspace, given all the configuration work I did there, especially related to email, is daunting.

Great, now this is in my brain. 🙂