I think it's a reasonable extrapolation.

Microsoft's historical strengths have been associated with enabling third parties.

Nokia, Surface, and xBox seemed to be them pining for a more Apple-like model where they just call all the shots up and down the stack.

Given that Nokia is dead, Surface is kind of de-emphasized, and xBox has started to see use as a brand for PC gaming, accessories, and partners... It's not a huge leap that the'll just delegate the brand to the OEMs on hardware execution as the software stack hasn't really benefitted from a locked down hardware BOM in quite some time. Microsoft played with in-house hardware and didn't seem to have particularly impressive results while sort of risking alienating their partners that have driven their strength.

Further there was a paper about how LLMs were able to beat intermediate chess playing humans.

Then I dug in and they would do things like allow an LLM several rounds to correct a mistaken move (e.g. the engine would just make up a new bishop or make a piece move in an invalid way). If they had just given the LLM one and only one shot at each move, who knows how many games would have failed.

AI is obnoxiously overused in marketing and invites skepticism immediately as it strikes people as lazy marketing trying to cache in on a blatantly hyped phrase. It starts to smell scammy when all the scammers are right in the bandwagon of using it.

Further, it doesn't talk about *what* the product does, but just says it uses AI to do it. People want to know what is good about a product, and wouldn't care if it's done by some credibly "AI" approach, or a traditional programming, or a breed of super intelligent hamsters operating the device remotely from some operation centers.

The point is 'new' and 'cutting-edge' are uselessly vague, but have long been a staple of marketing speak that people have pretty much tuned it out. They wanted a 'control' for a marketing word and those are pretty much best they could come up.

I would say the marketing issues aren't even about LLMs or any specific problematic experience, but being wary of an overused buzzword. It might have some further negative impression owing to the threat that AI is going to come for their jobs. I'd say it's a tiny fraction of people that actually have given an LLM a spin and found it less useful than everyone seems to act like it is, either they haven't bothered or they used it so lightly that it seemed vaguely credible.

Also quite a bit more awkward.

RH was on a trajectory to decently compete with VMware as a decent virtualization platform on technical merit, but I think they found VMware being first meant there wasn't much interest in changing.

So they ditched RHEV and chased "cloud" with open stack... Except open stack was never that great, and the demand for a fully realized on premise "cloud" didn't follow from off premise cloud anyway...

So red hat changed to openshift and kind of sort of shoehorned VMs awkwardly to provide some whiff of continuity to customers otherwise abandoned for trying to adopt RHEV or Open Stack from red hat.

Proxmox I think it's on the best position for providing the old fashioned VMware user experience now (in some ways better).

Another company whose business strategy requires customers believe they have a handle on this 'AI thing' claim they have a handle on this 'AI thing'.

Maybe they do, maybe they don't, but it's hardly a trustworthy source of truth rather than just doing their marketing work for them by relaying their marketing effort as 'news'.

Well, actually hardware consolidation is a use case, sure, but I think nowadays it's more about redundancy, fault tolerance, rapid deployment/decomissioning. If you are doing it right, your "OS" boot volumes should be considered disposable, but a lot of shops do it wrong and want the OS volumes to be hosted in centralized storage, which is much easier with a virtual machine approach (yes you can SAN/iSCSI boot, but it's not very appealing).

Of course your first point stands, that VMWare has competition with adequate capability and until now vmware could largely get by because the customers are too lazy to move and the price wasn't enough to make them look hard at options. It's not like mainframe style lock in where the porting effort is supremely daunting, though what you say about they don't need it to last too long for it to have been worth it also stands.

Of course, the nice meaty IBM locked in ecosystem is far stickier than VMWare.

That replacement for a mainframe, can it run exactly the same software? Generally not, it has to be ported, and porting is risk.

For VMWare, the replacements can run the same exact applications (the processor architecture and software stack have nothing to do with vmware's part of the solution). A customer may be *somewhat* stickier as they bought into vmware-centric solutions with partners, but as they migrate to newer hardware platforms, they can comfortably look at alternatives without terror that their applications are doomed if they try.

There's some friction against migration of course, but no where near what mainframe enjoys.

Reading into the article and his choice of words, sure some get some negotiated break, but the key is his use of the phrase:
complaints "don't play out"

What he means is that the customers complain, but the complaints are invalid because the customer isn't using as much of the feature set as they *could*, and so the complaint has no merit because they are getting what they paid for even if it's useless to them.

He then goes to either fabricate or cherry pick a few examples of what a customer might say when they recognize what fools they have been and how much more they could get out of their vmware purchase.

Well clearly you just need to start using 31.1k/year of the additional value you will be getting /s

Ubuntu LTS has "Pro" offerings that take it out that far, and Windows isn't free, so it seems fair to include their paid expanded support.

The reason I wouldn't use the RHEL/Alma/Rocky is that I am impatient for new features, but if I was a "I don't care I want to run this for 10 years", then I'd run it on my desktop. I think this is mostly the reason enthusiasts dislike them, which is an opposed concern to "not supported long enough". RHEL10 recently released based on Fedora 40, where desktop enthusiasts are running a Fedora edition a whole year newer.

For Fedora, the "click here to upgrade" is pretty similar to the Windows "click here to upgrade" experience. Unless you get adventurous in ways you couldn't have gotten adventurous in Windows.

As a Fedora user, sometimes you have a period of software instability when they push something not yet baked. It may be for a reason, but that reason may be nearly impossible to discern.

It's not news because the community is broadly used to it and they generally accept it as the cost of getting stuff faster.

Fedora is not as bad as it used to be, but they are really aggressive and inflict oddities from time to time.

If I were really bothered, I could go run something extra conservative, like Debian Stable or Alma Linux, but I prefer the fast-ish delivery of Fedora even accepting that sometimes things can go a bit south.

What LTS editions only do 5 years? I just checked SUSE, Ubuntu, and RHEL.

RedHat is up to 13 years, with the the first 5 years being "full" including releasing for brand new hardware and backporting as needed with another 5 years of "you can keep running it on the hardware you have, but we aren't promising support for new hardware" and another available 3 years of paid extension. Note that Windows 10 pretty much went "maintenance" with the release of Windows 11, so the RHEL lifecycle largely imitates the Windows lifecycle.

SUSE is a bit more generous on paper, but roughly this is about all the LTSes.

However day to day users are not interested and go for the options that favor rapid delivery of new capability, so people don't talk about them as much.

Exactly. Even if a technology might have a shot at being desirable, I often see seller interests trample the value and then the seller surprised that the customers didn't go for it after they did absolutely nothing to cater to the user base.

One company I worked at had this persistent issue and a strong warning sign was that they just absolutely worshipped the fictional Henry Ford quote about customers just wanting faster horses and the inventor knowing better than the customer about what the customer should want.

Even adjusting for "all movement is somewhat useful for the skill of driving", an AI model driving consumes training material way more than a human will ever see in their lifetime if they popped right out of a womb and drove for every waking and sleeping moment of their life, several times over. The amount of input and feedback about spacial navigation from just moving about is still a tiny amount by both amount of movement and hours of movement of the training data.

Same for text processing, not only does it consume more than a human will ever see, it will consume more text than a human will ever see, hear, conceptualize across many lifetimes.

Yes, the AI scenarios have a more narrow scope of material but the volume of it is still inordinately more than a human will consume no matter how much you credit somewhat different experiences as "equivalent".