Ask Paul: February 7 (Premium)

Happy Friday! It almost feels like a reckoning on multiple fronts this week, but there’s a lot going on, and many great questions. Let’s jump in.

? China, tariffs, and … Windows 10 EOL?

JustMe asks:

This one is a little out there, but it may make for some interesting discussion. What affect, if any, do you think the recent tarrifs on China announced by the US will have on Windows 10 end of life? My thought process was maybe a little, if the tarrifs cause chips to get more expensive thereby making new hardware more expensive and consumers (possibly small businesses as well) hold off on implementing any sort of hardware refresh/new PC buy until the prices come back down. I dont expect Microsoft to move the EOL date for W10, but I do wonder if this might catch a few people out.

Because of the volatile, unpredictable, and intellectually and morally challenged nature of the current administration, we don’t need to worry about the established economic realities of tariffs here. That is, on paper, these tariffs should be incredibly challenging for chipmakers like Intel, AMD, Qualcomm, and Nvidia, not to mention the innumerable companies that make other hardware components that go into PCs and other electronic devices. But that’s not how things work right now. Instead, there will be protectionist deals for U.S.-based companies, as there was for Apple previously. Of course, many of the components that go into PCs and devices aren’t made by U.S.-based companies, and so prices will still go up for the customers–consumers, businesses, whatever–that buy the resulting product.

Will this be enough to impact the Windows 10 end of life (EOL) schedule? I don’t see that happening. I was just looking at the Windows 10 extended security updates (ESU), and the costs for consumers ($30 for one year) and businesses ($45 to 61 per user for the first of three possible years), and feel like that’s the end of it. This EOL date was telegraphed well in advance, companies should have been plotting whatever migrations or upgrades well in advance, and any mainstream use with a PC that doesn’t qualify for Windows 11 by late 2025 or 2026 needs to upgrade, period.

It will be interesting to see the real-world impact on PC and device pricing. Given what’s happened so far, I expect a lot of bluster and then very little in the way of change. Especially if those changes were to raise prices for Americans.

But the Windows 10 EOL? That train left the station.

?️ Secret sauce

louiem3 asks:

Hi Paul, simple question but I have been having a hard time finding an answer in regard to Copilot Pro. Do you know what LLM model it uses? Some places I’ve read it uses GPT-4 along with its own models that Microsoft has developed but nothing concrete.

Microsoft will never give us the specifics of how it modifies the output provided by OpenAI’s models in Copilot, and there are rumors it will soon switch to a model in which Copilot could use different models, including those not made by OpenAI. But it has discussed what it does at a high level. It uses an orchestrator called Prometheus to prompt one of several OpenAI models–today, GPT-4, GPT-4 Turbo, and GPT4-o–and combine that with Bing indexing and search results and unstated refinements to produce answers. When it first announced Copilot, then called Bing Chat, it described this as its “secret sauce.” But the models it uses, and the methods it uses to arrive at answers, have of course evolved over the past two years.

I wrote a bit about this in A Shift in the AI Power Dynamic? (Premium) recently. Today, it’s interesting to compare the output of ChatGPT with that of Copilot when they are prompted identically because they use the same underlying models. But that may be even more interesting in the future, if and when Microsoft evolves Copilot to work with other (non-OpenAI) models. This may be a temporary condition, but different models each have whatever strengths, and the Prometheus orchestrator that powers Bing could emerge as an advantage of sorts if it has a range of models to choose from and can route queries correctly and in a sophisticated manner to get the best possible answers.

Tied to the point below, I mentioned to my wife, who recently used AI effectively for a work project and was surprised by how well it worked, that we’re in a temporary period in which different AIs are jockeying for our time and money. But ultimately, we’ll use what we use. If you’re in Office, it probably makes most sense to use Copilot, and if you’re in Google Docs, Gemini will be there, etc. These things will all work similarly. And it’s not difficult to imagine a future in which the big AI infrastructure companies are all cross-licensing each other’s models and all the models provided by the smaller AI companies too.

I’ve been using it more since I pay for O365 Family and have found it useful so far. Aside from using it as a stand-alone app in Windows, I also use it in Edge to summarize pages which is great for long articles and such.

This is what happens. It’s no longer possible to argue against the efficacy of AI in personal productivity, and while some continue to resist a technology that is so clearly beneficial and so broadly, this is still what happens. You will use it. You will have that eye-opening moment. You will use it more and more. It will just become a part of the fabric of life, an expectation, like electricity or connectivity. And we’ll wonder what we were debating for some reason for 15 seconds in the mid-2020s.

My wife’s ah-ha moment is perhaps instructive. Like me, she’s a writer and, like me, she doesn’t want to use AI for writing, beyond grammar and spell checking. (We mutually agree that all the AI-based writing tools that help with rewriting in various ways are incredibly useful to non-writers.) She spends a lot of time on the phone interviewing people, usually doctors and other medical professionals or people who have experienced some health-related issue. For the past several years, she has recorded those interviews with Zoom, transcribed them with the Transcribe feature in Microsoft Word (originally only in the web version), and then used that text to help construct the resulting article she writes.

If you think about that process, there’s time spent talking to people–and I hear that happening every day, literally–there’s an automated (and no doubt AI-based) transcription process, and then there’s time spent going through the transcriptions in which she pulls out specific points, quotes, or whatever. And then she finally does the thing that I would describe as her jobs: She writes.

The other day, I showed her how Click to Do in a coming version of Windows 11 will make any on-screen text or graphics interactive by letting you click on them and get a menu of actions you can perform. For text, those actions include rewriting tools. A coming version of Notepad has even more sophisticated rewriting tools, but the thing about Click to Do is that it’s happening at the OS level. And that means it will work in any app. I use a Markdown editor called Typora for writing, so I don’t have direct access to Copilot/Copilot for Microsoft 365/Copilot Pro. But because Click to Do works with any app, I could use those tools in Typora.

Inspired by this, my wife took the transcripts from a recent (and in-progress) project about foods that fight fatigue and asked Gemini and ChatGPT to summarize the foods that each person had discussed. She was impressed by how well organized that was, and by how much faster it was than doing this manually, given the volume of foods involved and the occasional overlaps. Each food has a corresponding reason why it (allegedly) fights fatigue, which is (or isn’t) a scientific/medical fact that can be verified, and many rely on specific preparations (for example, a fresh fruit may be beneficial where cooking it is less so, or whatever). So there’s a lot to organize there.

Next, she used these AIs to discover whether these claims could individually be proven. She asked the AIs to find evidence-based research from the past five years to support each claim, with citations. And then she confirmed whether those sources did, in fact, support the claims.

This saved her an enormous amount of time. Normally, she would have to do this research herself, and spend time (as I do with programming issues) on Google, whack-a-moling in and out of websites that may or may not be credible and may or may not have the answer she needs. (She has specific sites that are untrustworthy as sources and cannot ever be used, like Wikipedia.) She was blown away by how well this worked and by the accuracy, as she was able to verify each claim. And with that, she has shifted into a place where this is how she works. She will never go back, and that’s what I meant up front by, this is what happens. There is now. And there is the before times. And in the before times, we were less efficient.

(She is using the free versions of these tools. I told her she should keep trying a few different AIs and that, over time, she’ll start to feel that one or the other is, for whatever reason, better for her work. And that she should just keep using the free version until she can’t. At that point, if that AI is so useful to her daily work that she has to pay for it in whatever capacity, then she can certainly justify the monthly cost. But I wonder if that will happen.)

The important point here is that my wife, a writer, is using AI to get work done, though she is not using AI to write for her or to rewrite something she wrote (i.e. doing her job for her). She could do that, and as noted, that’s a great capability for non-writers. This, to me, is literally why AI exists: To get people past those things they are not good at or don’t understand at all, which enables them to do their actual work. In her case, again, there are four steps to the resulting articles she writes, and now AI has automated, to some degree, two of them, and neither is central to what she does for a living. Those things–transcribing calls and collating data points from different interviews–are just rote work, time-consuming tasks during which she is not writing. Now, most of her time can be spent on the calls, which is unavoidable, and–wait for it–on writing. Which is her job.

? The year of Linux on the desktop

AnOldAmigaUser asks:

My daily driver is running Windows 10, but I also have a laptop running Windows 11. I have been testing Linux, mainly because I have a surplus of devices that refuse to die, but cannot be upgraded to Windows 11. As a child of children of the depression, it goes against my nature to get rid of them as they work just fine, surprisingly well as a matter of fact with Linux. I am taking a course in Conceptual Furniture Design at a local college and the lab there uses Macs so at the moment I am regularly hopping between OSes.

There’s a lot of anxiety over PCs ending up in landfills tied to Windows 10 EOL, but those PCs were always going to end up in landfills. And I’m not sure whether it matters whether that happens 5, 10, or 20 years after the PC was built. This is a long-time issue that is perhaps better measured in geological time. And it’s not Microsoft’s “fault” because it only supports a specific Windows version for 10 years. Collectively, it’s all our faults. We need to figure out how to recycle that hardware and build more responsible products going forward.

But it doesn’t matter. The good news is that we have options for these PCs. And those that wish to keep using them can just keep doing so, using an out-of-support Windows 10 or Windows 11, using a workaround. And that’s been the case for a long time. I always wonder about the out-of-date Windows versions I see on PCs in doctor’s offices, in hotels, or wherever. But we’ve all seen that, and we always will.

Putting Linux–or perhaps Chrome OS Flex–on an older PC is an even better option. Linux is lightweight compared to Windows, of course, and there are even specific Linux distributions aimed at older PCs that are even more lightweight. So that can be a nearly ideal answer to the aging PC problem, with the small asterisk that Linux remains a bit technically challenged regarding user experience, hardware compatibility, and whatever else. It’s a lot better than it used to be, for sure.

Looking only at the OS, I would rank my preference as Linux first, then sort of a toss-up between Windows and MacOS. Linux just seems to get out of the way, whereas Windows can’t seem to get out of my face, and MacOS is just a bit quirky. There are certain applications that I use that mean I will continue to have a Windows device, but it may not be my daily driver. A few years ago, this would not have been the case.

I still prefer Windows, of course. I’m not sure whether I prefer macOS, Linux, or Chrome OS Flex, as that shifts from time-to-time. The Mac is obviously the most polished and mature, and the hardware is great. Linux and Chrome OS Flex both have good stories but speak to different audiences. It’s weird to me that a mainstream Linux distribution hasn’t really emerged, though I guess Ubuntu or maybe Mint are closest to that.

In reading the list of devices and applications in “What I Use – January 2025” I was struck by how few Microsoft applications you use. You have been an early and consistent voice against enshittification in Windows, starting with “ads/suggestions” in Windows 8. Yet I know, that you regularly test other platforms, and just as regularly state that you prefer Windows as an OS. Why? Obviously, it is a personal choice, but I did not see anything in What I Use that would tie you to one OS or another.

It’s easy to toss off my preference for Windows as familiarity, but that’s not exactly right. I mean, obviously. I am more familiar with Windows than other platforms. But I’ve never felt the need to make, say, the Mac work more like Windows per se. Instead, I try to be open to different ways of doing things, and I can identify specific things about, say, macOS, that I prefer or like less than similar UIs in Windows. For example, the touchpad gestures in macOS for whatever reason feel more natural and seamless than they do in Windows, which is weird because they’re nearly identical. But I very much prefer how basic task switching (Alt + Tab) works in Windows vs. the Mac because the latter handles separate windows in the same app differently (i.e. poorly) and that’s just inefficient. (Things like Stage Manager and full screen apps further complicates this.) So that’s not “familiarity,” it’s just an objective conclusion based on experience. One way is better than the other.

There’s no simple reason that I prefer Windows to the Mac or whatever else. There is instead a series of micro-reasons that add up to me preferring Windows. This is true of any big decision, I think. There are pros and cons. As an individual, we can weight those according to what matters most. And then we choose. If you think about this in terms of smartphones, I may prefer iOS to Android overall, or vice versa, it doesn’t matter. But one reality is that when I use iOS, I do miss certain things from Android (specifically Pixel Android). And when I use Android, I miss certain things from iOS.

When I use Windows, I don’t miss anything from macOS or whatever else. But I like how full-screen apps work on the Mac. I like the touchpad gestures. I like that Apple doesn’t enshittify the Mac with advertising or tracking or whatever else. (Apple does bog down all of its OS platforms with an astonishing number of apps, many of which are arguably in crapware territory, and I’m surprised there isn’t more attention to that. But you can uninstall all that.) The hardware is fantastic, functionally and aesthetically.

The non-Windows platforms all have what I think of as blockers. And if there are enough blockers, or if one thing I need is weighted so heavily that not having that thing makes it a blocker so serious I can’t use the platform, I have to move on. When I use Linux or Chrome OS Flex, or the Mac, I always bump into these things that give me pause. They can be big blockers–the lack of availability of a cloud service to local file system automatic sync feature like that provided by OneDrive or Google Drive–that are central to my workflow. Or they can be little things, annoyances that add up. Like me not being able to Alt/Cmd + Tab to every single window for some stupid reason.

Some blockers can be solved with utilities. I use a program called AltTab on the Mac to solve the aforementioned multitasking issue, for example. There are all kinds of third-party utilities that let you access OneDrive or Google Drive (or whatever other service) in Linux, but none are seamless as they are in Windows (or Mac). So I experiment. I do what I do. As I wrote this week in There’s Nothing Wrong with Notion, So I’m Messing with It (Premium), I am curiously open to replacing the tools I use the most often. That’s always been true with Windows. But I have not replaced it. It’s still the best choice. For me. We’re all different.

Coincidentally, I have been looking at Linux this past week. I’m in Mexico as I write this and so I have fewer PCs to use and test things with than I do back in Pennsylvania. So this isn’t the best time for me to be experimenting at the OS level. And yet, I have: I’ve been looking at Elementary OS, Zorin OS, Ubuntu, and Linux Mint on an HP laptop here in a dual-boot configuration with mixed results. Maybe this is just a compulsive thing on my part, but I spent a lot of time getting the laptop up-to-date and customized with Windows and then decided to screw with it. But whatever. Maybe someday something will change–Windows will keep getting worse, some other platform will improve in some way–and I will change too. Maybe my workflow will change, making other platforms more viable to me. (Per the Online Accounts 2025 (Premium) post, moving to a NAS-centric storage for work could alter that decision matrix.) I’m open to whatever.

In looking up this topic on my site, I came across Replacing Windows? (Premium) from seven years ago, and that post describes some of the testing I’ve done over the years of various Windows alternatives. I’ve been pushing on this for a long time.

? Photos

oasis21 asks:

I have a bunch of photos that I need to scan. I want them to be organized by date taken. Have you found a better way to set the Date Taken tag of pictures than the one you suggested in First Steps: Scan Old Photos and Back Them Up to the Cloud to use “Windows Photo Gallery”? Is that even available?

That post is from 2017. I tackled this more recently in my Digital Decluttering series in 2023/2024, and getting my photo collection organized was the biggest and most time-consuming part of that. I did “complete” that work, however, and I wasn’t sure that was even possible.

There is no “good” way to set the Date taken meta-data on photos per se, but I wrote about this topic specifically in Digital Decluttering: Tagging, Deduping, and Replicating the Photo Collection (Premium). It describes the apps I used to find image files without Date taken meta-data (or with incorrect Date taken meta-data), and the apps I used to bulk apply this meta-data when possible. (This was a huge problem with older, mostly scanned photos, and then much less of a problem when you get to newer, phone-based photos.) This is the type of thing where I was heavily involved, developed specific skills, and got pretty good at it. And then once it was done, I forgot how I did it, and will likely never need to do anything quite like that again. But there is a wrap-up, of sorts, in Digital Decluttering: Photo Consolidation by the Numbers (Premium) as well.

Good luck. This is a thankless task in some ways, but having finished it, I’m glad I did it.

?Customer service

train_wreck asks;

Have you recovered from the Google fallout, and have you received ANY indication of what might have triggered it? I didn’t need any more reasons to hate Google, but this certainly adds to the pile.

This much I know. I will never get any satisfaction, let alone any response, from Google on this. Ever.

I had forgotten about this–my key coping technique, literally, is to forget bad things so they don’t consume my brain–but this isn’t the first time that Google screwed me over like this. In 2023, Google quietly changed the additional storage options it offered to Workspace customers, so the 1 TB of additional storage I had been paying for on top of my Workspace subscription and relied on was taken away. Worse, Google informed me of this change only 30 days before the change, which would turn my 1 TB of individual storage into 30 GB of storage pooled between 5 users. Um. What.

In that case, I ended up moving all my photos from my Workspace (paul at thurrott.com) account to my personal Gmail account, and I now pay for a 2 TB Google One subscription there. But I had contacted Google Workspace support to see whether or how I could switch my Workspace subscription to a higher-end tier with storage, while leaving the other subscription at the Workspace Business Starter level with pooled storage. This would have been a lot easier than downloading my entire photo collection, deleting it, and then uploading it to Gmail. But my experience with that was similar to that with YouTube: With time marching forward, Google responded slowly to my many emails asking for help while literally threatening me with data loss. And in that case, some of the email interactions I had with support before they finally drifted off and stopped responding were nonsensical. Here’s one example.

I received this email 30 days after I asked for help. It was the second email from support. The first arrived 22 days after I submitted my request, and it was an automated email said that “a member of the support team is reviewing your ticket. You will receive a response from our team shortly.” That reply arrived 8 days later (as above). Two weeks after my confused response to the Vietnam thing, I got another reply. “It looks as though we do not have an account created for you yet within our Vector system,” it reads. “Can you please provide the following info?” The info they wanted was the name and domain name of my company. It helpfully promised to help me switch my one account to the higher subscription tier. I replied. I never heard back. I asked what was happening three weeks later (having long since moved to Google One). They asked me for the same info again. I replied with, “Never mind. It’s been over a month. You are less than worthless.”

This was the reply I received to that.

This was the last I heard from Google Workspace. It arrived over six weeks after I first contacted them for help regarding a matter that that involved a lot of data and needed to be resolved within 30 days per their schedule. As with YouTube, I emailed them repeatedly for help and was usually ignored.

Do I “hate” Google for this or whatever else? No. But these events are wake-up calls that remind me that I cannot rely on this or any other company to do right by me. I’m on my own and I need to look out for my best interests. And so that’s what I’m working on right now. To me, recovering from this will be me not relying on this or any other Big Tech company as much as is possible.

Thanks for the reminder, Google. I had let my guard down. I won’t make that mistake again.

? Dev focus

iantrem asks:

You mentioned today on FRD about the Build conference and I’d commented on Laurent’s post on Wednesday about it that it’ll just be a 3-day advert for Co-pilot. As I develop websites that are hosted on Azure but I’m not interested in adding AI functionality (and I don’t have any clients currently willing to pay for it), do you think they’ll be anything at the conference that won’t be AI based, or shall I just stick to the “What’s new in C#” talk and be done with it? What I’m basically asking is, is AI all that’s new in the world of Microsoft and development?

Yes. No, really. But yes.

Build is a big show, and there will obviously be sessions about .NET, Windows, web, and mobile development, and whatever else. But AI is the focus at Microsoft now, and so Build will focus on that as well. The question is whether the other content is enough to warrant attending.

This is a tough one. Generally and for me personally. Build is my favorite Microsoft conference and industry event. My focus is very much on Windows, of course, but I recognize that’s almost pointless from a dev perspective, though hope springs eternal from a news perspective. The cloud focus of the previous several years was tough for me (maybe better for you). And while Copilot/AI is somewhat interesting to me from a dev perspective, it’s also changing rapidly and is at best a side topic to me. And … I don’t know. Right now, I’m not sure if I’m going to Build myself.

The case against attending is obvious enough: It’s expensive and a time commitment, and most of the sessions and all the keynotes will be available on video replay anyway.

Hopefully, we’ll get a session catalog quickly enough that you can decide based on that. But Build sells out so quickly now it may not matter, and you may have to decide in a vacuum. I hate recommending that you not go, and there are good reasons to go regardless involving all the in-person/networking opportunities, especially with Microsoft employees who work on the products you are focused on. I’m not sure what to say.