ZFS sort of moved inside the NVMe controller - it also checksums and scrubs things all the time, you just don't see it. This does not, however, support multi-device redundant storage, but that is not a concern for Apple - the vast majority of their devices have only one storage device.
Strange specs table - it seems to ignore the tensor core FLOPs, which is what you'd be using most of the time if you're interested in computational throughput.
I think they can be, except of course the problems need to be much harder, and impossible to solve via vibe coding alone. Like it or not, AI assistance is going to stay with us. This is the "new baseline" against which engineers will be judged.
Hmm, generating royalty-free music on Suno, or getting involved with some pompous dude who strokes out at the sight of someone eating a burger and demands 51% for his "no strings attached" stuff that's 20 years old? Tough choice.
I woke up to this in 2016 when theretofore beloved public figure Donald Trump turned into literally Hitler immediately after he descended down that elevator in the Trump tower, all without changing a single opinion he'd ever held. And it's been unrelenting ever since.
Did I say it wasn't? If your context is short and your model is small, it is possible to run LLMs on high-end CPUs able to support 12 channels of high-spec DDR5 RDIMMs. It's not possible to run them as fast as they'd run on a GPU equipped with HBM though. Nor would it be even remotely as energy efficient. Also, it's not possible to run LLMs quickly on CPU if your context is long, because CPUs do not have the requisite FLOPS to process long context quickly. And before you bring MoE into the conversation, MoE only affects the feedforward part of each transformer block, and full memory bandwidth and compute savings are only realized at batch size 1, sequence length 1, AKA the most inefficient mode that nobody other than Ollama users use in practice. Sequence length 8 (common for speculative decoding) could be using up to 8x37B parameters (assuming you want to run DeepSeek - the strongest available open weights model). Batch size of even 2 with sequence length 8 could use almost all parameters if you're particularly unlucky. Prompt will almost certainly use all parameters, and will slam into the FLOPS wall of your EPYC's ALUs. So can LLMs (with an emphasis on "Large") be run on CPUs? Yes. Are you going to have a good time running them this way? No.
Note that this is _way_ slower at small batch sizes you'd need for interactive use. At batch size 1 this seems to run at 1/3rd the speed of bf16 (so about 1/6th the speed of fp8 you'd realistically be using) if figure 5 is to be believed. This is actually a pretty impressive feat in itself if you know anything about GPU kernel programming, but it is much slower nevertheless. For this to work at "wire speed" it'd need hardware support, which takes years. Their "baseline" elsewhere in the paper is CPU offloading, which is dog slow and can't be made fast due to PCIe bottleneck.
Original title: "Former New Mexico judge and wife arrested by ICE". It's as though he wasn't an active judge while harboring an alleged Tren De Aragua gang member.
Protip: believe absolutely nothing you read in mainstream news sources on any even remotely political topic. Read between the lines, sort of like people used to read Pravda in the Soviet Union.
And all the drugs and treatments derived from those "studies" are going to continue to be prescribed for another couple of decades, much like they were cutting people up to "cure ulcers" long after it was proven that an antibiotic is all you really need to cure it. It took about a decade for that bulletproof, 100% reproducible study to make much of a difference in the field.
As far as I can tell, it's "$20K" the same way Cybertruck was "$39K". It's not available for purchase yet, and when it is, it'll be twice as much, because Bezos also likes money.
Why though? Does anyone use it? I tried to use it on my iPad Pro and couldn't get used to it. This would make sense if some iPhone model was a tri-fold foldable with a huge screen, but I don't think that's going to happen in the foreseeable future.
It's kinda like Suno, except Suno sounds pretty good sometimes. Even so, I played with Suno for a few days and lost interest. There are some amazing examples on Suno, though: https://suno.com/song/9a7fd58e-132c-4ac5-9a25-f40d7f6f8c9f. This is one of the early tunes, it probably can do better now.
We'll see about that, won't we. Printing $2T in an average year and paying $1T+ in interest _with borrowed money_ certainly does feel like a Ponzi scheme to me. No way out of this either, only collapse, further underscoring the ponzi-like properties of the system. The question is only when.
Printing arbitrary amounts of money on a whim without actually backing it with anything does undermine the solvency of the government, however. As we will discover within the next decade or two, once US dollar loses its reserve status.
I went the other way around. I carry my Z8 everywhere (in the backpack or on the strap), take it with me when I travel anywhere, and lately I've been shooting quite a bit of video with it as well. I started doing that when I realized how little I remember from just 5 years ago, and how helpful apps like Immich or Google Photos are with remembering the fleeting moments of my life. Could I do it with my iPhone? Yes. But I find that having a device that _only_ takes photos and videos makes things more intentional and less distracting.
I thought this was a settled issue: the date is August 15, 1971. This is the exact date the US Government became a Ponzi scheme. On August 15 1971 gold was $35 per troy ounce. Today it's 3342. Not to worry though, all other governments are Ponzi schemes, too. We do not need to run faster than the bear - only faster than the next guy.
Looking at the comments, the blowback to this is pretty wonderful to see. There's hope for HN after all. More and more people are realizing that they don't have to blindly subscribe to the "current thing".
And then throw in some co-located compute there using chiplets, entirely bypassing the memory bus and PCIe. That'd be _the_ ideal Transformer substrate. Memory bandwidth bottleneck just disappears for the most part. Memory size bottleneck, too.
Someone at Google should sic an llm at the trove of documents from the Microsoft antitrust trial. This is directly from the 90s Microsoft playbook. I'm sure I'm by far not the only one with a sufficient attention span to remember that.
Typically only the most powerful models are worth a try and even then they feel like they aren't capable enough. This is not surprising: to the best of my knowledge none of the current SOTA models was trained to reason about 3D geometry. With Grok there's just one model: Grok3. With OpenAI I used o1 and o3 (after o3 was released). With Google, the visual feedback was with Gemini Pro 2.5. Deepseek also serves only one model. Where there is a toggle (Grok and Deepseek), "thinking" was enabled.
I've done this, and printed actual models AIs generated. In my experience Grok does the best job with this - it one shots even the more elaborate designs (with thinking). Gemini often screws up, but it sometimes can (get this!) figure things out if you show it what the errors are, as a screenshot. This in particular gives me hope that some kind of RL loop can be built around this. OpenAI models screw up and can't fix the errors (common symptom: generate slightly different model with the same exact flaws). DeepSeek is about at the same level at OpenSCAD as OpenAI. I have not tried Claude.
He's also not the only one doing it. Every time the Deep State needs to sneak in something particularly vile and avoid public scrutiny, we get "UFOs", or "directed energy weapons", or "Russian/Iranian/North Korean hackers" or something like that. "UFOs" have been used for this since the 60s. "Directed energy weapons"/"Cuban syndrome" - since the 90s. "Hackers" are more recent. Works like a charm.
I don't think they even wanted to be in the crypto mining market. They artificially hobbled their GPUs specifically so that they wouldn't perform well there.