Ask Paul: April 17 ⭐️

Happy Friday! We’re heading off to Cuernavaca for a long weekend away, so let’s get this going with a blessedly lighter-than-usual Ask Paul after last week’s magnum opus.

“I Ditched Password Generators and Made a Better One in Excel”

You have wasted your life.

💻 Marketshare is hard

helix2301 asks:

On MacBreak Weekly, they were talking about how Apple MacBook Neo sold way more than expected, and Apple is starting to run out of old chips. The argument was that Apple was going to use MacBook Neo to get a larger marketshare.

These things are difficult to know unless Apple provides any real data, but it’s clear that the MacBook Neo has sold well and you can see on Apple’s website that the delivery time for a purchase is out to two to three weeks. Among the unknowns is how many of the A18 Pro chips Apple has and/or can be manufactured and what the plans are for the future, with one rumor pointing to the A19 Pro and a 12 GB RAM configuration that would definitely be an improvement over the current 8 GB.

But stepping back and looking at this more generally, I’ve never understood why the Mac doesn’t have higher marketshare. This dates back, literally, to the 2001 Mac OS X launch, which was technically impressive and quickly became viable for mainstream audiences (say, by the time Tiger landed in 2005). Apple has always had impressive hardware designs, the prices were almost always on-point, with the mainstream configurations in a sort of entry-level premium part of the market. And in more recent years, the Mac has benefitted just from being in the Apple ecosystem where the iPhone halo effect is quite real. And yet … somehow, inexplicably, Apple has never moved the needle on Mac usage or marketshare numbers. I don’t get it.

So yes, the MacBook Neo is an interesting example of good timing, arriving as it did right as the component shortage was kneecapping PC makers already hobbled by low margins, and right pricing, something Apple doesn’t always get right. There was nothing inevitable about this product per se–if it had cost $100 to $200 more than it does, I’m not sure anyone would have complained or found that odd–but Apple really did “think different” here. Yes, it has low-end Apple Watch and iPhone devices, but neither is as hobbled as this thing is. And the timing, with iPadOS 26 finally turning that device into a real computer if you want that, is also curious. But Apple is curious. They often seem to zig when the rest of the world zags.

Then on FRD, you said Apple is 6th biggest PC manufactor and this big gap between 5 and 6, and a bigger gap when you get into the top 3.

If you look at the most recent quarterly PC sales data from IDC and the most recent annual (2025) data, you can see that Apple is the number four computer maker. It ended 2025 with about 9 percent marketshare–which is units sold–and one has to assume the Neo contributed to Apple hitting 9.5 percent share in the most recent quarter. It’s reasonable to wonder about the rest of the year and Apple finally breaking the 10 percent share milestone. I feel this is achievable. But again, I’m curious why this didn’t happen previously. Surely, MacBook Air is strong enough to help Apple achieve at least 10 percent overall. But it hasn’t been, somehow.

Do you think Apple has any chance to break into the top 3 as a PC manufacturer?

Yes. Obviously, all we need to do here is look at number three, which is Dell, and its share/units sold numbers and see where that’s going. And Dell is in an odd place. Most of the computers it sells are commercial (business) PCs, not consumer PCs. And the company’s overall focus has shifted dramatically from PCs and traditional servers to cloud-based datacenter/AI servers. And with Nvidia rumored to be looking to buy a PC maker, Dell’s PC business is perhaps the most obvious choice. (I recommend Microsoft Surface, but that’s another story.)

In 2025, Dell sold 41.1 million units (up 5.1 percent) and obtained a marketshare of 14.4 percent. Apple sold 25.6 million units (up 11.1 percent) and, again, had 9 percent marketshare. To overtake Dell, Apple would need to sell dramatically more Macs and Dell, well, would need to sell fewer units. But this is possible: Thanks to component shortages and price increases, it’s highly likely that businesses will extend their usage of existing PCs and stretch out the buying cycles. And that would play right into Apple’s hand, since its computers target consumers. And the MacBook Neo is the one bright spot, price-wise, in the industry.

I always think of these things in terms of simple numbers. If you’re watching the ending seconds of a basketball game and the score is close, with one team only up by 2 points, it’s fascinating how different things can get depending on what happens next: If the team with the ball hits a three-point shot, they’re up by 5. But if the other team steals the ball and scores three, they’re suddenly down by 1. Both outcomes are, sorry, game changers to some degree.

With that thinking, if Dell’s share fell 3 percent and Apple’s share rose 3 percent, things change pretty dramatically. Dell’s share would be roughly 11 percent and Apple’s would be 12 percent. Granted, that’s several million units in either direction for both companies. But that does feel possible. Even though going from 9 to 12 percent is actually a gigantic leap forward.

Is it likely? That I can’t say. The needle will likely move slowly, over some number of quarters. And at some point, the component shortage will ease up and end. Apple will release second and third generation Neos, and may need to raise the price as the specifications improve. Etc. But possible? Absolutely.

“Testing the MacBook Neo on 8K video and raw photos”

You have wasted your life. And our time.

🔮 Path to success

Will asks:

With Copilot now adding Claude as a model option and rolling out things like Cowork, I’m curious why we haven’t really seen Microsoft say much about its own models inside Copilot.

I think it’s because they’re not ready. This is one of those “Rome wasn’t built in a day” things, in this case because it’s not possible to just spin up a wholly new top-tier AI model from scratch that will instantly threaten the best that Anthropic, Google, and OpenAI have to offer.

But the market is also changing to some degree and while a big-bang LLM that can do everything and anything will probably always be a thing, big (and medium and small) AI models that are custom-tailored for specific needs and workloads will be much more common. (I guess they already are.) And so as Microsoft AI is building out its family of foundational models, it is benefitting from all kinds of things tied to the passage of time and experience. Had it started this work three years ago or whatever, it would have done things differently.

What we’re not seeing, in other words, is an MAI-1 model or whatever they might call it, a GPT alternative. Instead, we get specific models like MAI-Image-2, MAI-Voice-1, and MAI-Transcribe-1, and if you think about each in terms of the types of first-party capabilities Microsoft would offer to customers (through Microsoft 365 and/or 365 Copilot), you can see that they make sense for that. These are literally foundational to what it is that Microsoft does in the productivity space.

More generally, it’s pretty clear to me that a “Copilot”-branded AI whatever could be successful for Microsoft even if it just falls into the model (general model, not AI model) that Apple is doing, where there’s this brand (Copilot and Apple Intelligence) and these AI services throughout the ecosystem, but the models change under the covers as they evolve so that the company is automatically always providing the “best” ones as needed and is likewise letting customers sign in to whatever third-party AIs they use if they want that instead. This is basic infrastructure stuff, I guess, but Copilot, like Apple Intelligence, could evolve into an orchestrated set of AI services, if that makes sense.

I guess it does, based on …

It makes me wonder if Copilot is basically becoming a bundle of models and agents, with Microsoft choosing what responds to what. And if that’s the case, does that actually make Copilot better, or does it just create more confusion? We’re already seeing users unsure about which agent they’re supposed to use, and it feels like that problem could keep growing.

One of the many issues with AI right now is that we, as users or customers, are too aware of what’s happening under the covers. It’s weird to me that I could use whatever AI chatbot and have a choice of models that I manually select based on what it is I’m doing. I shouldn’t ever see that. I should just make my request and it should use the best model every time, best meaning whatever makes the most sense for that task. But we will definitely get there. We forget how early it still is, in part because the advances come so fast.

This automatic model selection, or orchestration, is available in some places and Microsoft is setting up its 365 Copilot to work that way now, which you can see in it mixing and matching OpenAI and Anthropic models in some cases. And soon, I think, Microsoft AI models. But the other way to look at this is that auto model selection is a form of simplification from the standpoint of users, as would be removing or hiding model selection UIs. For users, this is about “just do it.”

Microsoft is busy removing Copilot icons from various apps and other interfaces because of customer backlash, and it should be concerned that it’s undermined what was clearly a great brand. But there’s no end in sight to the AI capabilities it will add across the stack, especially in Microsoft 365. And in time, I think, much of it can or will be powered by Microsoft AI foundational models. It’s strategic to Microsoft that it compete in this space, just as it was in web browsers in the late 1990s.

A broader question for you, if you could give Microsoft advice on where to take Copilot next, what would you recommend?

To be clear, I’m not qualified in any way to opine on what Microsoft could do to make Copilot successful. I’m too busy driving my own small company into the ground and I have no particular business acumen. My advice would focus on meeting customer needs, and that may or may not be in Microsoft’s best interests here. It doesn’t help that Microsoft’s introduction of AI functionality has been chaotic and often feels directionless.

(To be fair to Microsoft, one criticism I normally have is that it moves too ponderously and sees everything as a big platform it needs to erect fully before it can release anything for customers. You see this in Loop, which is nothing more or less than a Notion clone with a byzantine platform layer built below it that may or may not ever work properly. With AI, at least, it has tried not to do that. It’s building a platform, definitely, but the Microsoft AI foundation model work came well after it started hurling AI features everywhere. In some ways, this is impressive, though the customer harm is problematic.)

I’ve never used the Anthropic plug-ins for Word, Excel, or PowerPoint, but I do know that the efficacy of those things has rattled the Microsoft 365 organization because it’s working on the same types of features but is moving much more slowly. Anthropic is a bit like Apple under Steve Jobs in the 2000s, meaning it looks at what needs to be done and just does that, and does it well and with differentiation, often creating things that its executives and employees want. Microsoft is like Microsoft has always been, plodding and platform focused, and it takes a while to get things out the door because all the scaffolding has to be made first. It should be more like Anthropic when it comes to AI, in short. Deliberate but fast. Figure out the top 5 or 10 customer needs in each use case and just do that first.

Microsoft is in some ways a victim of its success. Microsoft 365 dominates in personal productivity, but that dominance makes it difficult to think differently and potentially disrupt well-established (one might say “legacy”) products like Word, Excel, and PowerPoint. It has done so, of course, occasionally. Microsoft Teams is the best and most recent example, and the ribbon UI (which is closing in on being 20 years old somehow) is another. But when you’re a hammer, you see everything as a nail, so to speak. And I think there are major changes to how people are productive using computers or whatever devices. And companies like Microsoft that make the old thing are the least likely to deliver on something new.

When mobile disrupted personal computing, Microsoft belatedly responded by porting its core Office apps to the iPhone and then the iPad and Android. But lost to time and memory are these other efforts it made to make sense of mobile earlier on, things like Sway, that were small but also a different way of looking at a traditional workflow or use case. We’re on the cusp of a major disruption in how people produce content–”are productive,” if you will–just as we are undergoing now a major change in how we create software apps and services. Some will continue to be stuck in the old ways, just as some of my friends still prefer to send email instead of texting with WhatsApp or whatever. But innovation is occurring, and that will trigger disruption.

If you think about the creation of the iPhone, there were all kinds of things that went into that product occurring, but one of the big decision points was the iPod. Apple had made this incredibly successful MP3 player, and it dominated the market. But when it looked at what might disrupt the iPod, it realized that a smartphone was the obvious choice. And after a few missteps, it finally decided that it would disrupt its own product. So it created the iPhone. And that is what Microsoft should do with Copilot. Use it to disrupt Microsoft 365, not just be a rainbow-colored icon in all the traditional apps that, soon, only old-timers will use.

Obviously, this should happen alongside whatever improvements other teams are making to Microsoft 365. But it should happen.

The good news is that Microsoft will be fine no matter what happens. It doesn’t want to be just infrastructure, but that’s what AWS is and AWS is a fantastically successful business. It will try first- and third-party strategies, the first-party stuff being Copilot and Microsoft 365, and inertia will help keep those things going as-is. But disruption is coming. And it seems wiser to be the disruptor than to react belatedly to the changes. Have a say in that, lead in that. Don’t follow.

I guess none of that is particularly specific. However, I am working on a Premium post that goes into more detail about this disruption. Hopefully I can wrap that up soon.

“This private AI assistant runs offline for life”

Whose life? Also, you have wasted your life.