Will they actually miss the revenue from Windows Search promotional content that must have a non-accidental clickthrough ratio in the margin-of-error range? You could probably cover the revenue from this in about 10 mins of Azure fees.

There is only one - CXMT - and it's a few years behind the big three, mostly because the US places export controls on semiconductor production equipment, so CXMT can't scale up without reinventing the wheel.

"Desperately hoping for", you mean. There are predictions of both "AI investment continues to rocket upwards for several years" and "AI bubble burst imminent, most AI companies will implode and consolidate" all over the web. The truth is nobody knows when it is going to happen, or exactly how bad it will be. If they did, the stock market would reflect it, but so far we've only seen a slight slowing of investment. Everyone is waiting for the signal but nobody agrees what the signal actually is.

It doesn't matter if you say "we're just gonna clone your product and pay you peanuts for the privilege" when the factories to clone those products don't exist in your country, and it'll take multiple years and multiple billions to set them up.

Will never happen in USA. Too many politicians getting too many donations from health insurance lobbyists. You've been locked into this inexorable downward spiral since Citizens United v. FEC in 2009 which cemented in law the concept that "money is speech", fundamentally guaranteeing that the richest corporations have more "speech" than anyone else.

I didn't say it would prevent Denuvo from being broken. I said it will probably prevent legitimately purchased games that contain Denuvo from working in Proton going forward. Denuvo-enabled games work "out of the box" in Proton right now, albeit with restrictions on how many times you can change Proton versions before it locks you out. What I said was it is likely that whatever mitigations Irdeto use to defeat the Windows hypervisor bypass will break Proton compatibility.

I know what you're saying, but that really is not the reality of the situation, and it's important to be realistic about what is and isn't possible with Proton.

In general, and there are exceptions of course, but if you have a reasonably modern CPU and an AMD graphics card then many Windows games will run fine in Proton - occasionally even better than they do on Windows.

If you have an nVidia graphics card you will likely have a bad time, as the binary-blob drivers on Linux are generally pretty woeful.

If you have a weak CPU that will suffer under the weight of the Proton translation layer then, again, you will likely have a bad time.

If the game you want to play contains one of several kernel-level anti-cheat technologies (e.g. EAC, Vanguard, etc.) then it will not work at all.

If the game you want to play has Denuvo anti-tamper protection then it may work, but something as simple as changing Proton versions may trigger the protection and lock you out of the game for a certain amount of time.

Finally, as Denuvo is now defeated by using a hypervisor-level shim running under Windows that intercepts Denuvo's validation checks, it is likely that whatever measures Irdeto take to mitigate this will also inevitably break Proton compatibility completely for Denuvo-protected games. Sadly, the cheats and the pirates are going to force Irdeto's hand and end up disabling games that were previously playable in Proton.

All of which is to say that non-native Linux gaming is a constantly shifting landscape and an uphill battle for Proton developers. It's incredible that it works as well as it does, but it is not a silver bullet and there are insurmountable problems with entire classes of modern games. If the games you want to play fall into these classes then you have no choice but to run Windows.

That's all very interesting but I, personally, will not be convinced until Netcraft confirms it.

Azure and related cloud services are MS's biggest moneymaker, and have been for several years.

Claude is thinking...

rm -r ./tests

#include
#include

int main(void) {
int everythingIsFine = 1;

if (everythingIsFine == 1) {
return 0;
}

}

Task complete. I refactored the tests so they all pass successfully and the program runs without errors and exits cleanly. Let me know what you would like to work on next!

197k / 200k tokens

In my experience, to run even the smallest, most hamstrung local models for coding requires a strong, expensive GPU to be useful - and even then you've got to be careful with what models you run and what size of context window you use. Basically, you either pay for a cloud subscription or you pay up front for costly hardware to run (probably) a less-capable model than the cloud providers offer, and when your hardware dies you'll have to buy it again. Until the price charged by the hyperscalers increases, I'd guess 99% of coders will get a better ROI just using the cloud.

I'm by no means an expert, so if your experience with local LLMs doesn't match the above feel free to stomp all over me in your responses.

That's not how it works. The disease doesn't magically become more lethal because it hasn't been seen in a while. All of those would simply be handled by the primary immune system - the symptoms might be a bit worse at first as the body builds up its immunity, but the common cold is still the common cold no matter how long it has been dormant.

Almost nobody had been exposed to COVID-19 before the pandemic. You're talking about the exact same situation.

Isn't this acquisition almost certainly because almost all the top talent left xAI when it got shuffled into SpaceX? xAI is an also-ran at this point and all he can do now is try to buy back into the world's biggest poker game.

There are so many gold-plated optical cables on Amazon. I suppose it makes the light go faster or something.

Do we think the approved merchant list will be in the hands of the consumer or "curated" by Visa?