kryptiskt

Deciding on an ISA like RV32I only sets constraints on your architecture, it doesn't give you any of it for free. So I think it's entirely valid to say that it's a custom architecture if they didn't use any pre-existing CPU design.

Basically, they had the API interface and implemented everything behind it. I think saying that "it's an RV32I core" vastly underrates the design work that goes into actually implementing an RV32I core from scratch.

I think this largely is about brand protection. They worry that discounting the clothes means they will just cannibalize sales of that brand's full-price clothes.

I really wish I could buy the SpaceX part of SpaceX decoupled from all the AI hopium.

That's not for you to choose, and I'm glad someone is taking on the thankless task of correcting all your misinformation.

I'm a bit surprised that Red still seems stuck at 32 bits, like it was last time I saw something about it years ago. What is the roadblock that prevents them from building 64-bit binaries? I guess it must be very hard since otherwise it would have already been fixed.

There are no beefy RISC-V servers. There are no beefy 32-bit ARM servers.

It's a complete fantasy that Debian maintainers do a thorough review of changed packages. It's not a responsibility they have, and it would be impossible anyway (how many packages are there in Debian? How many maintainers are there?).

Typst does typesetting like TeX (or InDesign for a WYSIWYG alternative), neither org-mode nor markdown has a rich enough formatting language for general typesetting, like if you want to make a flyer for a concert, a brochure or a comic book.

Allow me to propose a modest alternative to space data centers, namely mountaintop data centers. This would consist of a container full of servers and GPUs and what else goes into a data center, a wind turbine for power and a communication module (say laser or microwave) for communicating with a base station with a fiber connection. This would be lifted on top of a mountain by a helicopter and bolted in place. Cooling would be provided by heat sinks exposed to the outside air. Some of the nodes could relay traffic from other nodes on remote mountain tops out of sight of the base station.

This scheme has many advantages over space data centers including launch costs, cooling, connection latency, servicability and ease of recycling.

If I want to identify and measure the size of round things in my orange sorter machine, I shouldn't have to resort to an unnecessarily complicated solution just because some AI bros can't understand that not everything needs to be an AI model.

Like, the AI model tools already exist, all that would be accomplished if OpenCV pivoted would be to take it away for people who want to do low-level vision programming. It wouldn't add anything useful to the world, just destroy an excellent library.

They fire everybody and then they bring in way cheaper European developers.

If the slaver is a human, other humans will judge them by human standards. Keeping the slaves ignorant of their condition and alternatives should not make it any more acceptable (if for nothing else since that is something you could easily replicate with human slaves by raising them as slaves).

Why would using math in certain fashion such that it mimics consciousness be considered unethical and comparable to human slavery?

If it is really conscious, it should have rights. Why? Because it's a person, with thoughts and experiences, and we're not evil and deprive persons of their right to self-determination because it's convenient to us.

I wouldn't be so pessimistic, Intel and AMD aren't going to stop making CPUs, and if their integrated graphics adds AV2 it will be motivation enough for others to follow.

That it's not possible for him to love her back as much as she does. "Requite" is quite an obscure word, I've only ever seen it used in the phrase "unrequited love", which means a love which isn't returned (in quite a different sense than what is used here, since I assume that the author didn't mean that he didn't love his wife, only that his love didn't measure up).

The certificate issuer doesn't have access to the underlying private keys, so while getting a fake certificate may be useful for MITM [0], undermining the certificate authorities doesn't actually allow spying on traffic that uses the genuine certs, no matter how corrupt the CA is.

There is such a thing as overestimating the power of the NSA, if the spooks actually had undermined the system to that degree they wouldn't need to lobby for all the surveillance bills that keeps popping up.

[0] And you can't get a fake certificate either without it being visible in the certificate transparency logs, or being an obvious fake since it is absent in those logs.

That Musk couldn't even get over the first stumbling block which is the statute of limitations, does not make the win any lesser. It makes it more decisive, since Musk now has to overcome that hurdle before even having a shot at the meat of the case.

Last commit was 2 minutes ago. Seems like a better measure than releases, different projects have different release cadences.

This doesn't compare like for like, since its comparing the total cost for the local machine with the usage cost for the cloud service, despite the cloud service also needing a local machine to be useful.

This reads like Apple fanfiction to me.

But then Apple can negotiate on another basis and say, well, if you don’t do us a favor here and give us a better rate, then maybe we won’t work with you when all this settles down. You know things are going to settle down. These things are always cyclical. There’s never been a semiconductor boom that’s not followed by a semiconductor bust. Never. And they know it.

I have to think that the RAM suppliers wouldn't be that easy to intimidate with threats, since they know perfectly well how few alternatives Apple has. And they are also perfectly aware that Apple will play hardball with them when the market turns, regardless of whether they were nice to Apple now.

It's not possible to learn anything about other elements when performing binary search, _except_ the only thing there is to learn: if the target is before or after the recently compared element.

You have another piece of information, you don't only know if the element was before or after the compared element. You can also know the delta between what you looked at and what you're looking for. And you also have the delta from the previous item you looked at.

You can to vary the split of the output by cracking heavier hydrocarbons into lighter. So it's not a fixed fraction, but driven by both demand and cost of processing.

I find all those arguments unconvincing. The right 10,000 lines of code can be worth a billion dollars. The idea that it would be somehow uneconomical for me to take the time to get it right feels like utter nonsense. I don't have to have much of an edge over an LLM to come out on top once you start to distribute the resulting product. Three months of my time costs $25,000 or so (hey, I'm in Europe, adjust as you see fit), if I can make something just a little bit better than AI Albert who can whip something together for a tenth of the price, my time will pay for itself once you have modest amounts of revenue from it.

And I'm fully convinced that what I do will not just be a little bit better than what AI Al makes. It will trounce it in all quality criteria. But of course, coincidentally with the rise of AI assistance, software quality has completely disappeared from the conversation. I wonder why.

Unbelievable, literally trillions upon trillions spent on all other kind of shit like warfare and utter corruption. And some people just can't get over that people fifty years ago spent a minuscule fraction of that to go to the moon.

You're positing the existence of a far more advanced lifeform than merely a clever monkey with pretensions, which then somehow created said monkeys. That's like saying that it's easy to become a millionaire, just start with a billion dollars.

That's not an explanation, you just replaced a problem with another harder one.

Taking care of yourself is not a competition, nor a zero-sum game.

But you don't get the plain, cold, hard truth in the second case. You just get an LLM with output in that style. The model will still be as path dependent as ever, it doesn't output the truest answer, it selects the answer that best fits the prompt.

Datacenter capex decreasing means that the chips have to go somewhere else, so it doesn't matter too much that the fab capacity has been spoken for, if the demand side is slacking prices will decrease.

I wouldn't call it optimized, since that implies that it gains performance due to the tail calls and would work otherwise, but the tail calls are integral to the function of the interpreter. It simply wouldn't work if the compiler can't be forced to emit them.

You lose in versatility, then you can't add user-defined operators, which is pretty easy with a Pratt parser.

The Web is a side project of CERN, they should have gotten a comped top-level domain by rights.