Android

It doesn't solve Safety Net/Play Integrity, at all. My bank is the kind that just warns you and then lets you in anyway. I just live without Google Pay, I just put the card in the phone case to the same effect. The point I was making there is that most apps don't care, Google isn't "pushing" it, but it is made available to developers, so really it's the app developers' choice to check or not.

Pixels are just less fiddling because flashing it is supported. It is not endorsed by Google, and you don't pass Play Integrity at all, but it is supported and doesn't void your warranty. They just allow you to install whatever you want on your hardware without a fuss, and get the full performance you'd expect and all, and even make use of the security chip. But, they only trust their code and their ROM for the purposes of Play Integrity, which is kinda fair game.

That's why it is quite ironically the device of choice for GrapheneOS. It's not a hack, it's a fully supported use case even though you lose Play Integrity certification, so they can implement all the security features Google has access to. The TEE will happily sign a unique and verifiable integrity attestation... for GrapheneOS's ROM signature. You can make an app that only works on genuine official GrapheneOS the same other apps do with Play Integrity. You can have a custom ROM and properly enroll it in some enterprise MDM and all that stuff, and only allow your builds of that custom ROM to enroll. But, no Play Integrity because it's not their official certified build.

It's like PC, you can turn off secure boot, you can secure boot with your own OS keys and get all the security benefits. But Valorant will still refuse to let you play if you haven't booted with secure boot into an official unmodified copy of Windows where they can ensure their kernel anti-cheat can trust the kernel about what drivers and processes are loaded. Microsoft isn't forcing their OS on you, but the developers will only trust you if you do. You're still perfectly free to put Linux on it, and it won't affect you otherwise.

If you're buying something to mod I'd recommend a Pixel, unless you're getting an older OnePlus for cheap.

You caught me. I meant this, but was thinking backwards from the bottom up. Like building the logic and registers required to satisfy the CISC instruction.

This mental space is my thar be dragons and wizards space on the edge of my comprehension and curiosity. The pipelines involved to execute a complex instruction like AVX loading a 512 bit word, while two logical cores are multi threading with cache prediction, along with the DRAM bus width limitations, to run tensor maths – are baffling to me.

I barely understood the Chips and Cheese article explaining how the primary bottleneck for running LLMs on a CPU is the L2 to L1 cache bus throughput. Conceptually that makes sense, but thinking in terms of the actual hardware, I can't answer, "why aren't AI models packaged and processed in blocks specifically sized for this cache bus limitation". If my cache bus is the limiting factor, duel threading for logical cores seems like asinine stupidity that poisons the cache. Or why an OS CPU scheduler is not equip to automatically detect or flag tensor math and isolate threads from kernel interrupts is beyond me.

Adding a layer to that and saying all of this is RISC cosplaying as CISC is my mental party clown cum serial killer... "but... but... it is 1 instruction..."