Ask Paul: May 9 (Premium)

Happy Friday! We’re back in Pennsylvania after almost four months, and the transition is going about as well as can be expected. We appreciate the clean air, the rain, and the greenness of Spring here, but we also miss our more active lifestyle from Mexico City. Nothing is ever perfect, I guess. So let’s kick off this weekend a bit early with some great reader questions. Next stop, an overdue music night.

? Beside a Surface

jrzoomer asks:

Hi Paul I noticed Microsoft announced new Surface products and to be honest I had totally forgot about the Surface brand.

COME ON MAN. 🙂

They’ve had four major product announcements in the past year!

• Surface Laptop 7 and Surface Pro 11 (Snapdragon X), May 2024
• Surface Laptop 7 and Surface Pro 11 for Business (Snapdragon X), September 2024
• Surface Laptop 7 and Surface Pro 11 for Business (Intel), January 2025
• Surface Laptop 13-inch and Surface Pro 12-inch for Business (Snapdragon X Plus), May 2025

It got me thinking more about the history of Microsoft’s hardware ventures and I wanted to pick your brain on one of my favorite ones, one that maybe informed the Surface line’s development and emphasis on design, but was also was a spectacular failure, the Zune.

What if Microsoft had never exited the consumer hardware market with the Zune? Was the loss to iPod inevitable? Could it have priced it better, or hired better designers? And ultimately I wonder if it could have launched a Surface like device before the iPad, and have a stronger consumer electronics brand?

Microsoft has a surprisingly long history with hardware. It’s first hardware product was a CP/M card for the Apple II series that was, at the time, the biggest market for CP/M. It bundled its first mouse with Word for MS-DOS. Etc. But by the time we got into the 2000s, Microsoft was losing its grip on consumers thanks to more consumer-focused companies, most notably Apple. (But also Amazon and Google.) With the iPod, I was just coincidentally (re)reading the internal emails from Microsoft from this time period, where they admitted that they had gotten “smoked” by the iPods and iTunes. And were worried that they would never catch up.

They were right to worry. The iPod happened at exactly the wrong time for Microsoft. It was still very much Windows-focused, and that meant partner-focused, and so it used the PC market strategy wherever it could. (Oddly, not with Xbox.) So long before there was a Zune, Microsoft had various Windows Media Audio (WMA)-based initiatives, eventually landing on PlaysForSure, and many device partners. But it was clunky. DRM-protected media was incompatible across services. It eventually created its own in-house service, MSN Music, which was honestly really cool for the day, after its top-level MTV partnership on a service called URGE, didn’t turn into anything special.

Microsoft tried to get WMA on the iPod, and Apple actually agreed to that, but the terms were so bad that Microsoft was forced to decline the offer. Nothing it did with partners worked against the iPod and its seamless integration, and so Microsoft did the unthinkable and created Zune. This was unique for it at the time because it was the hardware, the software, and the service, all integrated and unilateral, with no opportunity for partners anywhere. Apple-like, in other words.

Because this was so different, Microsoft screwed up the initial launch by trying to woo mainstream press, publications like People, Men’s Health, Cosmopolitan, and whatever else, and more directly take on Apple. But these publications thought of Microsoft as a fuddy-duddy from the past, and uncool, and that backfired badly. It eventually went crawling back to the traditional Microsoft-focused tech press and bloggers.

Functionally, Microsoft tried to differentiate Zune in various ways. But it was always obviously an iPod rip-off, a me-too product, and of little interest to potential customers. Everyone just wanted an iPod.

There was no way that a Microsoft music service, product, or technology was ever going to succeed against the iPod. It just didn’t understand that at the time. But it became obvious enough, and quickly, and by the time the admittedly excellent Zune 3 came out, it was too late.

Some good came out of this. The Zune used a user interface that was based on earlier work for Media Center and Portable Media Center, and it continued forward as Metro for Windows Phone 7 Series and then Windows 8. There are even elements of this in the Windows 10 Start menu and elsewhere. But as we transitioned into Windows 11 and that new user experience, that era is finally winding down. We’re onto a new user interface style finally.

The iPad is a bit trickier. Apple shocked everyone in the industry with the iPhone and most of the market struggled to adapt to the all-touch smartphone design, and then failed completely. Only Android survived, in part because Google was just getting into the market at the time and adapted quickly. (And in part because of its Microsoft-like partnership model that undercut Windows Phone/Mobile dramatically, making it more attractive.) The Windows 8/RT team’s idea that a tablet and a PC could be one device was actually a good one. And it got close with Surface RT in many ways. But the apps platform was immature and needed more time, and it was rushed to market in what I think of as an incomplete state. And by that time, it wasn’t enough to be there, you had to outdo Apple in some way. Surface RT was too compromised to be competitive, and Microsoft’s PC partners were still upset about Surface and didn’t want to adopt the underlying platform.

I think this all boils down to too little, too late, but to be fair to Microsoft, the world had changed, too. It stuck to the model that had worked for too long, not realizing that no longer worked, and it went after consumers that were then looking to other companies for better experiences. Windows was correctly seen a unreliable, so why would anyone trust that company for consumer devices, software, and services? Not that it really rose to the challenge anyway. It’s not like we missed out on anything by Zune not being so good that anyone would even consider buying one.

Related to this, helix2301 asks:

Does the NFL still use surface? I have heard nothing about that in while wondering if that’s still a thing and maybe that’s why Microsoft is keeping it around I know you talked about surface this week and on FRD.

Yes, all the coaches and players on the sidelines use Surface exclusively, so you see the blue and white Surface logo everywhere during NFL games.

? Hypocritical or ironic?

helix2301asks:

I got a question from someone, and I thought you might know the answer or have an idea. Microsoft is allowing consumers to pay $30 dollars a year for security updates on the consumer side for the first time. Do you think this is a security thing or another revenue generator for them or both. I can imagine it cost man hours to do updates so I was wondering what you thought on this?

It’s easy to be cynical here, but I view this as more of a positive. From Microsoft’s perspective, I suspect this is mostly about nullifying the obvious complaint otherwise.

It’s certainly unique in that Microsoft has done this type of thing at least twice, with Windows XP and 7, but only for businesses. And the consumer price is half as expensive as the business version, while only being offered for a single year (where businesses get up to three years at escalating price points). I don’t know what the cost of supporting Window 10 is per se, but $30 seems reasonable to me, and it will give people time to figure out the next steps.

It will be a while before this shows up, but I just recorded an episode of Hands-On Windows that runs down the options for those still using Windows 10, which boils down to:

• Do nothing. Not recommended, but if you’re using an up-to-date web browser and possibly a third-party AV solution, viable for the short term only and only on PCs that are light-use.
• Pay for one year of Extended Security Updates for Windows 10. Again, $30 is reasonable if you need time.
• Upgrade to Window 11. With a workaround if required, this is generally safe, though using a PC with an 8th-gen or old Intel (or comparable AMD) processor will be mostly lousy.
• Switch to Linux. It will run better, assuming the hardware is compatible. Whether it will run the software you use will vary greatly by use case. But the good news is that it’s easy to test Linux: You can run most distro right off the USB install disk and see how well it works before installing it. (You could also try Chrome OS Flex.)
• Buy a new computer. This could be a Windows PC, of course, ideally a Copilot+ PC, but also a Mac or Chromebook.

Three of those choices are free. One is inexpensive. None, arguably, are ideal. But it has been ten years.

? An LLM is like a box of chocolates

wright_is asks:

There was a report of a problem with the new Gemini model released this week, that it has turned into a prude and won’t answer questions or accept information of a risqué nature.

That’s reminiscent of the issue that Open AI had with ChatGPT recently, too, where the latest models were a little too sycophantic, so it pulled them. I was just playing around with Copilot Vision, and one of the things that’s immediately obvious in interacting with these things is that it’s so easy to anthropomorphize them. And that feels related to the issue you note, and to the ChatGPT issue. We think of these chatbots as people as we converse with them.

This has caused problems for some models, which have turned off protections for harassment, sexual content etc. the model ignores the settings and refused to answer questions, telling the people that are entering the prompts that such things shouldn’t be discussed etc.

Sounds good, on the face of it. The problem?

The systems that are using the model help sexual assault victims and r4p3 victims (hopefully gets around OpenWeb’s censorship) and they are using a system that helps them produce police reports and legal documents to document their case, another was for PTSD and depression patients to discuss their inner feelings, only to be told that such discussions are inappropriate and people shouldn’t discus such things. Those sorts of answers suddenly being chucked out are counter productive and harmful to the people using those systems.

In the past such systems would be thoroughly tested before they are released to users and any changes go through a big suite of tests, yet we are now in a situation with cloud based services, where the back ends of these systems are being replaced willy-nilly with new versions, which seem not to have been tested.

Shouldn’t such applications be using their own LLMs, where they have control over what changes are made and a new model can be tested properly, before it goes into production?

So, on one level, yes. Obviously, yes.

But also, tied to the discussion from last week, this is all happening in real-time, and it’s moving quickly. The story you reference is a very specific use case–a company building a platform for sexual assault survivors, rape victims, and so on–and I can’t imagine Google ignoring this issue, given that use case. I can’t speak to the platform this organization is building, of course, but the natural outcome here is that the model makers or third parties will adapt LLMs specifically to their needs, and already area. And as the underlying models evolve, and so quickly, things will break.

I keep coming back to the same thought here: Just as my doctor might reference Google or whatever online source to see whether two different medications can be safely taken together, a person dealing with mental health concerns, or abusive victims, or whatever, will use similar sources and now AI as a helper. Anyone who relies on such a thing solely is behaving unethically and potentially illegally. We need to understand where we are in time.

We seem to have become so reliant on hosted services over which we have no control, that we have been ignoring the dangers for the perceived benefits of a cheap service – it is cheap because we have no control over it, because it is rolled out at scale.

Sure. This is true with so many things these days. You can fix a bug in an online service immediately, which is good, because all users benefit immediately. But you can also just as easily introduce a bug.

Maybe these startups should bite the bullet, if they are working in such sensitive areas, that they are using their own models. There is also the problem of where the information goes that the users of these systems are going, are they being used by Gemini to train itself? Are their data leaks? It is traumatic enough to have been assaulted, but to then have your graphic description of the event available in a publicly facing LLM to be called up by voyeurs is another level of violation…

This is the competitive problem, and the source of Google’s initial problems two years ago: If you put on the brakes to fix things before deploying, another model provider will not, and they may get the business you want. We often talk about how the law can’t keep up with technology, and while that’s often not the case, this is a good example of that being true. We need a sort of basic human rights regulation.

In the meantime, any company that’s going to use AI for this sort of thing is going to need to be more careful. This is an ideal way to be sued into oblivion.

? Sweet spot, 2025 edition

spacecamel asks:

I have a new AMD 13 inch FrameWork Laptop coming in a couple of weeks. One of the things that I need to do is decide how much RAM and storage I need to start with. Considering that I will push the machine some with COD and other games when I am away from my XBOX, what do you think the sweet spot is for specs today? While I will be in a good spot where I can upgrade the RAM and storage later, I would prefer to not have to spend money too many times if I can help it.

First, congrats on the Framework, especially for getting an AMD version.

Last year, I celebrated the move to 16 GB of RAM as the proper minimum for most people. But there’s no version of this story where 8 GB is enough for some class of users. 16 GB in the minimum.

For gaming, I will argue that 32 GB is more appropriate, not so much as a minimum but as the sweet spot, the “proper” amount for videogames on a non-gaming laptop. I do play the latest Call of Duty on a variety of laptops, some of which have just 16 GB of RAM, and the experience is OK. If this isn’t your only way to play games, that may work. But I would personally get 32 GB.

Given the modular nature of this laptop, perhaps it could be configured with a single 16 GB RAM card, and then you could add a second later if that’s found to be lacking for your needs. I’m not sure if this varies across different AMD configurations, but what I’m looking says that a single 16 GB RAM card is $80, and there’s no discount for getting two. I would have a hard time not getting 32 GB.

Did I mention 32 GB? 🙂