What I'm getting at is there will be modeling, lighting, and animating "by hand" like before.

Not that this would be the best example, but one of these systems should be able
to be shown a picture of some lake in a park, you can ask for a Mario and Master Chief scenario
and entities that act expectedly Mario and Halo-like appear in world, exhibiting their default behavior.

How long would it take to build that in Unity or Unreal?

Yeah, everyone knows that times a million. That's not the point.
You have 100,000 PCs, each with a hot GPU vs. AI gen-ed from a load balanced server farm someplace.

All things being equal, and they're not, the AI generated stream experience would be superior,
infinitely customizable and scalable and reactive in ways beyond the design studio world's plans.

And contrary to some opinions, probably using way less power and way less dedicated hardware.
Like I said, we're not there yet, but it's possible and we've seen it.
It's just a matter of scale, mostly at this point.

What will a 'Game Studio' be?
The era of parametric shapes and vertices and triangles and shaders will soon be done and gone.
Game views, and the scenarios themselves will be generated on the fly, by request,
rendered by some AI engine, to meet one's specific, custom view requirements.
Music and voice too. It will be the way it's all delivered.

Not yet, but soon. A way easier path to hyper-realism than having to model and light it all,
and as we've seen, that they can't do it today without requiring an Nvidia space heater.

Maybe they've all been chat bots for a long time. Although the McConnell bot seems to be broken.

I'm confused why you're confused why I think that. Of course they would not have put in more lighting if everything was sodium.

Wow I'm glad you have it all figured out and have good luck. Regardless of type they are engineered to fail. Big Clive has confirmed that over and over again his analysis of LED lighting. I've lost all kinds of LED lights not just the bulb style. But yeah you're right. I'm doing it all wrong!

Where I live I doubt anyone saved much money going to LED street lights. For the first few years LEDs burned out quite often. Then with the power savings, they put in way more street lighting (part of why cities are getting brighter and brighter). Personally I much prefer sodium lighting. It's so much easier on my eyes.

Around the home, LEDs are things I love to hate. They are just engineered to fail and will never last the 20 years some claimed. I am constantly replacing LED screw-in bulbs. Sometimes the COBs themselves fail because they are over driven. Often the potted driver fails. Have I really saved any money with LEDs? Who knows.

Not easily. That's why I use the Floorp fork. They preserve the ability to set the UI to something sane.

Granted prices have gone up by 50% in the last year or two.

$1000? I don't think so. Beelink sells lots of windows 11 mini computers with 16 GB ram and pitiful SSDs (256-512) for under $300 USD. These computers run Linux quite well. I have several in my home lab. For me they work so much better than Pis for non-embedded, server-type work.

If you want truly open source hardware then RISCV is the future.

Unfortunately this is a common misconception. RISCV is not open source in any way that slashdotters recognize. When they say RISC V is "open" all it means is that you can implement a chip design with the RISC V ISA without a license. That's literally the only practical difference between it and ARM.

If you want to use an existing RISC V core in your own custom chip and don't want to make your own core, all cores available now are fully proprietary and must be licensed. Furthermore all the important bits such as the GPU are completely proprietary. So you still need proprietary firmware blobs. Also most cores implement extensions to the ISA which are proprietary.

Cranking out microcontrollers for pennies is neat but it doesn't make it open source.

From an end user point of view, and also Linux support, there's no advantage to RISC V over ARM. They both really suck, frankly.

For me it's been the year of the Linux desktop for about 30 years now. Shrug. Been on KDE Plasma 6 for the last two years and it works for me. I use Mate Desktop on my laptop and it's fine. Never bothered to learn Gnome 3 so I can't comment on it, other than to say I really hate client-side decorations and controls in the area where the window titlebar should be. I guess that's also why I don't use macOS or Windows anymore either.

Back in 1998 I first tried Linux on my PC and tried the various desktop environments of the day including FVWM95, and they all sucked really, and my terminal foo wasn't so good back then. KDE 1.1 was what made Linux finally work for me as my primary OS. I could finally ditch Windows 95. I wasn't too thrilled about the single click thing but it didn't take too long for KDE to introduce optional double clicking. The other killer app on Linux for me at the time was the KDevelop IDE. It gave me a real path from VS on Windows 2000 to using Linux for all my development. I no longer use KDevelop, but I still use kdbg from time to time.

My uni's CS department had a bunch of Solaris labs running on x86 hardware, and CDE. I found a bunch of KDE binary tarballs for Solaris and ran it from my home directory and it was really nice. Other students saw that and started running it too. Eventually the Sysadmins installed it for everyone. I think it really did change how students looked at Unix. Later on as a sysadmin myself I oversaw the complete replacement of HPUX and Solaris with Linux, and we gave students their choice of Gnome 2 or KDE. Both were well liked by the students.

Feeling old is exactly the point. My Linux desktop, whether it's Mate, Cinnamon, or KDE, is always configured to look and act exactly the same as it has for the last 30 years. I figure a successful major distro update is successful if the final result looks exactly the same as before the update.

I've been running Gnome 2 for decades now. The Mate Desktop is the direct descendant of Gnome 2. All the major distros have Mate spins of it. I happen to like Mate better than Cinnamon.

But I know run KDE Plasma and have it set up to look just as my Linux desktop has looked for 25 years.

Circle has stalled/died. It was originally a proof of concept of "Safe C++" which was proposed to the C++ standards committee. However they decided to go for "Profiles" instead. So Circle hasn't seen any development since then. Hopefully someone will pick it up. It's a good concept, and in my mind where C++ should have gone.