There’s some key details to not forget.

Factorio essentially kickstarted the genre. Satisfactory was inspired by it. I totally dig what Satisfactory has done but having a blueprint that is proven to work is skipping a lot of risk.

There is an inherent tradeoff between graphics and gameplay. Both have good reasons to focus on. Factorio has optimized it’s graphics and logic to an insane degree. Far beyond what is typically expected of an AAA game. You just don’t see that directly, since it provides value by absence. The game doesn’t even start to slow down until you are hundreds or thousands of hours in.

There is a reason AAA games frequently run badly even on top tier hardware, it’s because they prioritize graphical fidelity over all else. Optimization is often an afterthought, since programmers are expensive, and optimization doesn’t provide the immediately apparent value that graphics or new features do. Factorio had to take that risk though, because the game would not be fun if it couldn’t scale past the first ten hours.

Highly detailed graphics are very skillfully produced as well, but it’s a misunderstanding that a game’s code cannot be of similar quality and depth. A sort of graphical AAA vs functional AAA. Factorio took a lot of highly skilled programmers to pull off, while a graphically intensive game put those resources into their artists.

Let alone the massive backlog of previous generation games and emulators that you don’t need a top line PC for.

This ten times. It’s why the online discourse around AI is often so one sided. Anyone walking into a room where people are all nodding along to the same shallow, unnuanced statements, and throwing stones at anyone that points that out or tries to share their own contradictory to the group’s experience, even doing so in complete good faith, isn’t going to engage for long. And so that discussion is never going to turn nuanced since all the people interested in that have been ousted.

And it sucks, because there are real harms in AI that must be guarded from for which we need widespread support. But the hostility and closed minded discussions just causes people to tune out and contrarily be more open minded towards AI as a response to the closed mindedness.

I mean, your post says “Forcing windows down Xbox gamers throats”. So people that are already in the locked down Xbox ecosystem, not people that already know that and avoid Xbox. Xbox is just Windows but locked down. It’s both Microsoft.

I completely agree people should just go for a PC, but if someone was buying Xbox already, they aren’t in the mindset. So if they’re going to buy Xbox anyways, having a device that’s not locked down to some console OS means they can switch at any time and rid themselves of Xbox, since it’s your device. To me that’s definitely an improvement over the status quo, even if other better options were already available.

The contracts for the Steam Machine were already locked in before RAM and GPU shortage even started, which means they will (like other consoles) be able to provide a reserved amount of devices at a fixed and lower price. But likewise, this also means that for any console that did not have contracts locked in before shit went down, will suffer massively from this. Thus looking at current prices for hardware isn’t indicative of how much prices will rise. Steam Machine could be the most affordable gaming device of the next decade.

But Valve also isn’t stupid. PC has the unique position where it has one of the longest backlogs of backwards compatible games and applications. Which means you don’t need top of the line hardware. People are still gaming on 10 year old PCs, so the hardware for those will be much more affordable and still be able to play most games. Especially indie games with their insane price to performance to quality ratios. If top line gaming on PC becomes economically unviable, it will simply move down a notch.

You realize PC is not just Windows, right? It’s Windows, Linux, Mac.

Windows might be the most used PC platform but game developer are well aware attaching their success to Windows is not the right move. And that’s exactly the kind of freedom PC gives that consoles do not.

Honestly, asset flips or pure ai slop are often not something you would consider a ‘real game’. They are closer to a scam than anything. And they certainly wouldn’t be published by a reputable publisher. I think that’s also what OP was referring to, a game that meets some minimal level of development and involvement.

There’s truth to that currently yeah. I think my points still stand as well though, and in the long run you will still be out worse even if the upfront cost is currently cheaper. It also just seems contrary to the free and open nature of Linux, but if you don’t already own something you can upgrade and are currently strapped for cash, fair enough. But that’s also not going to change if you sink 400-500 dollars into it and that’s your budget for the next 5 years.

Or just… don’t buy consoles at all. Buy a mini PC (which you can upgrade too) or wait for the Steam Cube? Which would both be cheaper in the long run. Because why still funnel money into a company that seems to be adamant that it owns that machine (and lets be honest, could try and use any kind of kill switch or safeguard to stop you from doing so) and will wield your money as a weapon against you.

It’s like soliciting a stalker because you enjoy receiving random gifts in the mail with totally no strings attached.

Hate to say it, but you might be missing out on something you won’t ever be able to experience again afterwards. It’s like with episodic releases of TV shows, half the fun is sitting with friends discussing and overthinking what just happened while you wait for the next episode. Being there too long after community wide revelations, you can’t experience that head space of mystery and surprise again. Deltarune handles the episodic releases very well honestly, I’d understand if it was a series of bad partial releases.

You are probably confusing fine tuning with training. You can fine tune an existing model to produce more output in line with sample images, essentially embedding a default “style” into every thing it produces afterwards (Eg. LoRAs). That can be done with such a small image size, but it still requires the full model that was trained on likely billions of images.