I like the idea of framework but I think I've made it pretty clear in this topic that I don't swing either way on it since I've never owned one. I just have heard good things about it from people who have used/owned one.

And yes, AI is a gimmick. At least, running it locally is. Don't get me wrong: when applied properly it's incredible. But right now it's all in the cloud, so how powerful your computer is is entirely irrelevant, and at the glacial pace we're going in model minimization, it won't matter for a while. I call it glacial because that's exactly what it is: we see more and more powerful models coming out (Claude just beat ChatGPT on Chatbot Arena) but... They aren't models you could ever run locally. More and more powerful models and more and more demand for insanely huge GPU clusters. That's what we're seeing. Oh, Google has Gemini Nano, which (supposedly) can run on a phone, but I highly doubt it's all that accurate compared to GPT-4 or Claude Opus. The power of your computer won't matter until we either figure out how to prevent these models from requiring tens of thousands of GPUs just to run remotely efficiently or somehow we magically manage to fit the power of those clusters into a phone or laptop/desktop case. And that's assuming that these AI companies don't get forcefully shut down by government because they're gobbling up 4-8 percent of the US's power grid, which is quickly running dry because these companies are just opening up more and more data centers instead of waiting for us to have the capacity to transmit that kind of power first. Well, either that, or governments start passing stupid legislation that makes running an AI impossible without paying billions/trillions in intellectual property licensing fees because somehow training an AI model is "copyright infringement".  (Maybe, just maybe, this last point might not apply to highly specialized models for things like gene/material synthesis, but we haven't seen the last of these kinds of court cases on LLMs, and we won't unless governments start shutting down copyright maximalists, hard.)
Edit: another thing we need to solve is the training problem/building AI from scratch problem. Right now you practically have to be a domain expert. If we want AI to really take off, people need to be able to easily create them even for specialized tasks without needing to have PHDs in AI/ML/neural networks/etc.
Edit2: oh, and the "I like to hallucinate" problem. That's another big one. As is the "figure out what's AI-generated and what isn't" problem, though for that one throwing more AI isn't going to solve the solution... Is there even a solution to that last one?

76, agreed on swappable laptops. Who actually enjoys tinkering and swapping out laptop components?  I myself really don't, and that marketing is totally irrelevant to me and not worth any extra cash. And the socket for CPUs do change, sure you get a new board, but now you need new ram, SSD etc. When doing that, why not just buy a new laptop?
As for reviews, the first couple days reviews mean nothing. But the reviews that are fully featured after a couple months are good I think.  My experience, any defect in a laptop if it is major manufacturing defect will surface in the first 12 months to 24 months, the warranty period. If you don't have major problems after that,  then great. I had my Toshiba Satallite L50A battery always showed 0% charge after less than a year.  Then in 2016 to 2017, two years after I started using it, fan started to grind and got progressively louder.  I took it to the Toshiba shop  and paid good money to swap out the battery and fan.  Both worked fine until I got a new laptop. But a battery or fan on a computer shouldn't fail so early, and this was a midrange laptop.  My Eluktronics all the components are orderable online, so if fans fail, I can just order a new one and put it in, or keycaps. I broke a keycap, just ordered a new one and pressed it in.
Regarding AI models and performance though, does CPU performance increases matter though if software doesn't take advantage of it?  I  mean if performance went up 200%, but nothing takes advantage of that, and if everything already happens instantly, it won't really provide quality of life improvements.

@76, You do realize that not everyone has the luxury of being able to get a desktop right?
It's not just for monitary reasons either. Some people don't have the space.

The 250GB and 1TB Expansion Cards give you the performance of an internal drive with the flexibility of an external one. With a USB 3.2 Gen 2 interface, the 1TB card exceeds 1000 MB/s read and write speeds, while the 250GB card reaches 1000 MB/s read and 375 MB/s write speeds. Both are fast enough to run apps and even boot an operating system from, and you can plug them into other computers for high-speed file transfer.
Specs
add
Weight: 8g Dimensions: 38mm x 7mm x 30mm Phison U17 Flash controller Micron N28 NAND Exceeds 1000 MB/s read speeds Exceeds 1000 MB/s write speeds 50% post-consumer-recycled aluminum 30% post-consumer-recycled plastic

Yeah... they're really hiding stuff here lol.

It's not even really about tinkering. It's about making smart and informed finantial decisions. Is that a Framework or System 76 computer? Honestly I’d need to look more into it to know for sure. But I do know that buying a cheap computer every 6 months, a year, 2 years, because they break that often? That is not smart. Buying a thousand dollar Mac Book just to throw that out in a couple years because you can't replace the battery? That's not smart either, and really neither was I for buying one. Computers with swapable parts are more than just, "Oo, look at me, I can build!" It's more practical finance wize. I’m not the brightest person on the forum, but I do know that putting something together yourself is much cheaper than paying someone to do the labor for you. It also presents less opertunities for you to get screwed.

@86, the problem with your argument is that repairing a Mac is not exactly easy, either in the skills department or financially. Apple still is trying to make repairability as difficult as possible, meaning that very few of the mac's components these days are actually things you can repair/replace. Keyboard? Yep. Screen? Yep. Battery? Yep. Trackpad? Maybe. Anything else? Probably not (it's probably soldered). And unless something has changed, Apple still changes you $80 or more for their repair kit, which from what it sounds is really, really complicated and has lots of terms and conditions tacked on top.
I push System76 because I'm familiar with them and they're repairable, not because they come with Linux by default/run Linux really well. Framework sounds neat but I've never tried it. There's also Tuxedo, which is slightly cheaper than System76 and, from what it sounds, quite good. Both are user-serviceable brands, with Tuxedo's latest model being all AMD which is very repairable and is an absolute beast (though with the caveat that the battery life you get is only 3-6 hours or so). Still, I would think that anyone who wants a new computer would want it to be serviceable if only because sometimes, you don't always want to (or need to) send it back to the manufacturer to get repaired and just need to replace something simple like the Keyboard, local repair shops may not even know about the brand of computer you have (but if they do you probably don't want to pay the $300-$800 repair fee they'll throw at you or you don't want to go giving them your login credentials when they don't need it (some repair shops are notorious for asking for these when it's unnecessary, and it's usually so they can possibly go digging around in your files, it's quite creepy)), etc.

@mohamed
If you aren't planning to use the Mac as a Mac and just want to put Windows on it somehow, why bother?

Like I don't get what you're asking or trying to say still because either you want to use OS X for something in which case why are we discussing this, or you want to run Windows on it, in which case you've got to do a VM.

And to everyone else if I had to guess even midrange laptops are good for at least 3 years.

@dan
At the end of the day there's a big conflict between replaceable components and efficient/cheap.  People aren't soldering everything because we don't want you to be able to repair it, they're doing that because modern tech is too small and fragile to work otherwise.  It is actually a limitation, not something we do to screw you over.

@90
I can't say anything in response because you are asking me to compare pizza and cars or something.  If you want to run OS X get Mac. Otherwise, don't.  I can't say anything, they're fundamentally different options.

If you are asking about architectures more generally, we have kinda solved X86 emulation on Arm.  It's not as good as X86 on X86 because it's not going to support all the optional extensions, but it's not an interpreter and does basically recompile your program.  Both Windows and OS X can do it, and I'm sure Linux has one by now too.

Phones are all pretty much Arm, lots of embedded stuff that's big enough that they want to run Linux, Mac of course, there are Arm Windows laptops on the market.  AWS offers Arm servers and charges less than X86 ones for it, Google also sells them to you.  For a vast majority of software you support Arm by toggling an option and rebuilding with no code changes.  Most modern software is higher level languages anyway, and anyone doing cross-platform development already has to support both architectures.  It'll take a while, you shouldn't buy Arm right now unless it's Apple, and it is possible that Intel or AMD will surprise us.  But as things stand X86 is losing market share and there's lots of reasons to believe that will continue, not least of which being that people can license the Arm architecture and build their own chips to meet their specific needs.  Given the choice of cheaper and longer battery life while not losing performance, or going X86, I know which laptop I'd buy.  Workstations will be the last to fall, but as Apple has shown SoC Arm workstations are indeed quite viable.

Risc-v probably isn't going to take off in the PC/server market.  The primary driver of that isn't tech, it's social.  Arm has a chance because it was able to bootstrap to significant market share, and most people at the level of manufacturing chips don't care about openness or anything like that.  It has a chance, but it will take a very long time to prove itself sufficiently to get Microsoft on board for example, and of course Apple makes all their own chips and chose Arm.  Probably their biggest chance was if China chose them, but China seems to be going Loongson instead and seems to have enough leverage or whatever that everyone is going along with that, e.g. Rust and C/C++ compilers, etc. are all supporting it.