• A New Mexico judge has ordered Meta to pay $942 million under a law that so far as I can tell is both unconstitutionally vague and entirely inapplicable. (WSJ / MSN)
  
  Appeals lawyers making bank again.
  
  
  
• Meanwhile OpenAI has fired back at Apple saying that if they were able to steal all of Apple's trade secrets then they weren't all that secret, were they? (Tech Crunch)
  
  I'm not sure that is how it works, Sam.

• AirLLM is a new Python library that streams data on demand from SSD to VRAM for AI tasks, allowing you to run a 2.8 trillion parameter model like Kimi K3 in just 4GB of memory, at a speedy 292 seconds per token. (GitHub)
  
  So only about 10,000 times slower than what is considered acceptable performance.
  
  
  
• China has invented a new self-repairing dress made of living fungus. (Dezeen)
  
  No.
  
  
  
• Measuring the Pareto Frontier of AI models. (DeepSWE)
  
  A Pareto Frontier is a line drawn on a chart through a set of solutions that are each the best at a different task, but there is no single winner.
  
  An interesting point is that GPT 5.6 Luna - which I have never tried - is nearly as good as Claude Fable but much, much cheaper, with the total for the test suite coming to $0.61 vs. Fable's $21.63.
  
  It's even more jarring when compared to Anthropic's supposedly cheapest model, Sonnet, which is weaker than Luna and cost $26.40 to run the suite.
  
  
  
• TSMC is sitting on a billion dollars worth of Apple A20 CPUs waiting for DRAM chips to pair them with. (WCCFTech)
  
  Might want to take a seat. You could be here a while.
  
  
  
• Lenovo laptops in China have been spotted shipping with memory from TWSC (no relation). (Notebook Check)
  
  TWSC is not in the Big Three. It's not even in the Big Five, counting CXMT and Nanya. I'm not sure where it ranks.
  
  
  
• Elon Musk's Terafab is looking to hire a memory process engineer. (WCCFTech)
  
  They'd be silly if they weren't.

Posted by: larryw208 at Saturday, August 08 2026 02:34 AM (IJlDu)

Posted by: PatBuckman at Saturday, August 08 2026 03:01 AM (6eVrQ)

Hide Comments | Add Comment

• Cloudflare has announced its new open-source operating system, Cloudflare OS. (Phoronix)
  
  It is not an operating system.
  
  It's an application.
  
  Written in JavaScript.
  
  That depends on Cloudflare services to function.
  
  So... Thanks, I guess?

• China's CXMT aims to have 30% of global DRAM market share by 2030. (Tom's Hardware)
  
  I have plenty of DDR5 (and DDR4) memory and DDR6 isn't expected to arrive meaningfully on the desktop until... Well, AMD's Zen 7 architecture has been confirmed to support DDR5 and that will arrive at the start of 2029, so a couple of years after that.
  
  Go for it, guys.
  
  
  
• Microsoft absolutely never said that they recommended 32GB of RAM for gaming on Windows even though it has been transparently obvious that is what you have needed for several years. (Tom's Hardware)
  
  My old laptop was better in every way than this one except that it had 16GB of soldered RAM. Oh, and only 1TB of SSD. I never upgraded that. Which reminds me that I have four 1TB SSDs lying around that I can add to my... Five. Five 1TB SSDs.
  
  
  
• Nikita Bier is stepping down as head of X at X, after a year of...  Results. (Tech Crunch)
  
  I don't think he'll be particularly mixed.
  
  
  
• Meta's Muse Spark is joining the increasingly crowded AI coding space. (CNBC)
  
  Okay. Good for them.
  
  
  
• Apple's Private Relay isn't. (404 Media)
  
  Didn't I already run a story on this?
  
  
  
• Only you can make book by helping stamp out forest fires. (Ars Technica)
  
  I'll take a quinella on fire or no fire.
  

Musical Interlude

1 Musical Interlude - probably no a cappella version of this one either

Posted by: Frank at Thursday, August 06 2026 08:59 PM (9Oo74)

Posted by: PatBuckman at Thursday, August 06 2026 11:36 PM (6eVrQ)

Posted by: Rick C at Friday, August 07 2026 12:40 AM (yc+bG)

Posted by: Robin Munn at Saturday, August 08 2026 12:27 PM (PkD3w)

Hide Comments | Add Comment

• On the fifth of August of this year, SpaceX will blow up the Moon.  (Yahoo)
  
  Wait, that was...  Two hours ago.  Well, can't see the Moon right now.  Guess it worked.
  
  
  
• SpaceX has doubled its revenue since the same quarter last year.  (Tech Crunch)
  
  And more importantly, halved its losses.
  
  SpongMonkeys hardest hit.

• I'm convinced AI image generators act retarded to get you to burn usage credits.
  
  
  
• Can Reddit fend of a new wave of AI SEO spam?  (The Verge)
  
  No.
  
  Do I care?
  
  Also no.
  
  
  
• Apple says ex-employees have been stuffing classified documents down their pants and handing them over to OpenAI.  (Tech Crunch)
  
  Anthropic hardest hit.
  
  
  
• AMD's datacenter revenue doubled over last year, but gaming revenue declined by 30%.  (Tom's Hardware)
  
  Given that the game consoles are engaged in a mutual suicide pact and nobody is building new gaming PCs right now, I'm impressed that they held the decline to 30%.
  
  Overall revenues were up by 50%.
  
  
  

Musical Interlude

• Who is legally liable for Anthropic and OpenAI accidentally murdering eleven trillion people? It's complicated. (Tech Crunch)
  
  Oops, sorry, that was next week's question. Though the answer is the same as this week.
  
  Because corpses can't sue.
  
  
  
• Meanwhile Apple finally fixed Siri, only to find nobody cares. (Tech Crunch)
  
  Sudo make me a sandwich.
  

• After a record quarter, Palantir CEO correctly labelled the AI industry Marxist. (Tech Crunch)
  
  For the next quarterly results he will recite pi to a hundred decimal places.
  
  
  
• Only about 10% of memory sold at retail is DDR4. (WCCFTech)
  
  This despite it being significantly cheaper than DDR5.
  
  The advantage of DDR4 comes from the fact that you already have it, and don't need to buy it.
  
  
  
• Prices on Nvidia graphics cards continue to climb. (Tom's Hardware)
  
  The RTX 5090 now costs around $5090.
  
  
  
• Can you still play computer games on 4GB graphics cards in 2026? (Tom's Hardware)
  
  Fortunately, yes.
  

Posted by: PatBuckman at Tuesday, August 04 2026 10:57 PM (6eVrQ)

Posted by: Rick C at Tuesday, August 04 2026 11:16 PM (EABKx)

Hide Comments | Add Comment

• The Xbox may keep disc support as Sony shoots itself and its customers in the feet. (WCCFTech)
  
  This would be an intelligent and obvious move meaning there is no chance it will happen.
  
  
  
• Xbox has been hit with worldwide price increases. (WCCFTech)
  
  I bought one a while back. On the one hand, it has since increased in price by nearly 50%. On the other hand, it's sitting in my closet.

• AMD's new Zen 6 CPUs offer (reads article) piddly little improvements that it is impossible to get excited about. (Tom's Hardware)
  
  Though this is on top of them being about twice as fast as Zen 4 or Zen 5 desktop CPUs, so they still look pretty good. Though Zen 5 itself was nothing spectacular compared with Zen 4.
  
  
  
• New York has published new rules that will force online services to... Publish feeds showing only accounts users have subscribed to, in strict chronological order. (NY)
  
  The problem is that this new law is only in effect for minors.
  
  
  
• A group of teenage hikers relied on Google maps and ended up needing a rescue. (SFGate)

  When a group of teenagers set out to hike British Columbia’s famed Howe Sound Crest Trail in early July, Google Maps suggested the trek would take about five hours.
  
  Instead, the hike stretched deep into the night, leaving the group exhausted, dehydrated, injured, deeply stupid, and in need of a helicopter rescue.

  Uh, yeah. That trail involves CLIMBING A FUCKING MOUNTAIN.
  
  
  
• There's a shortage of MacBook Airs because of the shortage of memory. (Tech Crunch)
  
  Everybody hardest hit.

Noodling

This one turned out well. It's very non-obvious what she is, which is exactly how it works in the story.

• Microsoft has vowed to make Windows 11 actually work in 8GB of RAM, which is timely given that the company is shipping Windows PCs with 8GB of RAM.  (Tom's Hardware)
  
  Before the DRAM Apocalypse advice was simple: Do not buy a system with 8GB of RAM unless it is user-upgradable and you plan to immediately upgrade it.
  
  Now the advice is more along the lines of: Buy a Mac, or run Linux.
  
  Since both of those options hit Microsoft below the water line, the company is finally bothering to start to unshit its bed.  Long overdue and the outcome remains to be seen, but at least they're trying.  Or claiming to try.

• Scanning 7.6 petabytes of AI data for secrets.  (Truffle Security)
  
  This is the entire public repository of AI models at Hugging Face.
  
  The scan found 221,303 live credentials, $3.67 in change, and a Werther's Original that is presumed to still be edible.
  
  
  
• Is this Billboard Hot 100 hit AI slop?  (The Verge)
  
  (shake shake)
  
  Signs point to yeah, baby.
  
  
  
• Major record labels are proposing to join forces to force AI slop out of the charts.  (The Verge)
  
  Replacing it with regular slop.
  
  
  
• Apple's stock price has slid by 10% as the company notices that hey, they also need to pay more for memory.  (Yahoo)
  
  Apple does have long-term agreements in place, but those agreements aren't all that long term, and as they sign new agreements during the ongoing drought they find that they no longer have a decisive upper hand.
  
  
  
• Google plans to exempt Android developers in sanctioned nations from otherwise required verification.  (Ars Technica)
  
  So if you're a hacker based in North Korea, Cuba, or the Russian-occupied regions of Ukraine, it's business as normal.
  
  That makes sense.
  
  
  

Posted by: Robin Munn at Sunday, August 02 2026 09:54 PM (PkD3w)

Posted by: normal at Monday, August 03 2026 03:54 AM (7jZjl)

Hide Comments | Add Comment

• VC-backed startups commit more fraud and researchers think they know why. (Tech Crunch)
  
  Yes, it's exactly the reason you immediately thought of.
  
  Congratulations, you are now a researcher.

• Single DIMM gaming works better than you might think. (Tom's Hardware)
  
  Assuming you're not using integrated graphics. If you are, it works exactly as you would think.
  
  But if you have a graphics card, and the graphics card has enough memory of its own, and your single DIMM has enough memory, and you have an AMD X3D CPU, the performance loss is only around 3% on average.
  
  And given that I have a 128GB kit (2x64GB) and a 96GB kit (2x48GB) I could run four systems rather than two if they all had X3D processors. Not that I need four systems, but I could.
  
  
  
• Setting up OpenClaw isn't as easy as the internet claims. (Tom's Hardware)

  Having 48GB of available video memory should allow us to run a fairly powerful AI model on the system.

  Wait, 48GB of VRAM? What hardware... Oh, a Beelink mini PC, and that's shared memory with the integrated graphics.
  
  
  
• Tailscale didn't stop the Hugging Face intrusion - but it should have. (Tailscale)
  
  Interesting writeup on why the dog didn't bark in the night.
  
  
  
• AMD's new Radeon 9050 graphics card is poop. (WCCFTech)
  
  Not as poop as the RX 6400 I just bought, until you take price and power consumption into account, in which case it's more poop.
  
  
  
• Sony understands that its customers are upset about its plan to end disc support on PlayStation. (WCCFTech)
  
  It understands. It doesn't care. It's doing it anyway. Because fuck you that's why.
  
  
  

Noodling

• Renesas (a Japanese company formed by the merger of the semiconductor operations of Hitachi and Mitsubishi) has announced multiplexers for third-generation MRDIMMs.  (WCCFTech)
  
  You'd be entirely justified in asking Wut? at this point, but this is a great technology landing at precisely the wrong time.
  
  MRDIMMs are registered memory taken to the next level.  Since the memory modules have interface chips - the registers - on them, you can have multiple banks of memory on each module, allowing 256GB DIMMs instead of just 64GB.
  
  What the M part stands for is multiplexed, meaning that when you access the memory it interleaves data from two banks at once, so that transfers are twice as fast.
  
  The new multiplexers run at 16GHz.
  
  Which means you get user upgradable high-capacity memory modules that run at the sort of speed normally associated with the memory on video cards.
  
  Which would be amazing and allow for some seriously powerful integrated graphics except that memory is so expensive right now that there is absolutely no point to it.

• LLMs are broken and can't ever be fixed.  (GitHub)
  
  We've touched on this before.  Computers have separate areas in memory reserved for instructions and data respectively, with strict controls over which is which.  When those controls break down, you get a common class of security vulnerabilities - buffer overflows and the like.
  
  LLMs don't have that.  Worse, they pass benchmarks running tests for that problem with flying colours - typically blocking 100% of attacks - but when faced with a real human attacker they fail 100% of the time.
  
  
  
• What we know of Nvidia's 5000 Super series.  (WCCFTech)
  
  Supposedly they are already in production but have been hidden away because nobody knows how to price them.
  
  
  
• Also, I bought an RX 6400 today.  The least powerful recent AMD graphics card and by a good margin, but also the least power hungry, and physically the smallest.  It's not much slower than the RX 580 (I have a couple of those) but is a tiny low profile card that only uses 50W of power.
  
  Also a lot cheaper and smaller than, say, an RTX 3050, while not being that much slower.  Not something you'd buy unless you need to fit the card in a low-profile slot, but the Hyte / Hololive limited edition cases (of which I now have five) only allow for a single full-height card, so I do.
  
  
  
• 2x, not 10x.  (OBryant)
  
  Sounds about right.

• The United States government just banned Roombas.  (The Verge)
  
  Well, it didn't, and wouldn't have at all if it hadn't previously blocked the acquisition of US company iRobot by US company Amazon leaving the market to Chinese competitors.  And they're only being blocked over well-founded fears of being used for surveillance just like every other thing in our lives these days with more functional purpose than a Chia Pet.
  
  If you like your Roomba you can keep your Roomba, and existing models of Roomba can continue to be imported, but new models of Roomba - and of any other imported robot - will need to be certified.

• Huawei's new MateBook Pro S weighs just 798 grams, which is vanishingly light for a 14" model.  (Notebook Check)
  
  Only problem is that it runs an in-house Kirin XE90 Arm CPU, which has decent multi-threaded performance thanks to its 10 dual-thread cores, but pretty lousy single-threaded performance.  Expect performance like a 2017-era laptop in that regard.
  
  
  
• Speaking of which, I just realised that the new Minisforum NAB9S has a 13900HK CPU rather than a 13900HX.  So while it is a current model that takes DDR4 SODIMMs (of which I have an abundance) and is a fair bit faster than my existing hardware that uses that memory, it is also a fair bit slower than the full HX model.
  
  It's also relatively inexpensive because it's available as a barebones model, and has a 2.5" bay for a SATA SSD if I should happen to find where mine have gone.
  
  
  
• How to run Gemma 26B on an 8GB Apple Silicon system.  (GitHub)
  
  Gemma 26B requires 52GB of VRAM to run as it ships with 16-bit values.  The process squimshes it down to 4-bit quantisation and then caches the core (using about 2GB of RAM in total) and streams the rest from SSD.
  
  Which is great because while you can certainly run Gemma 26B-A4B (4-bit quantised) on a 16GB graphics card without hacks, this sort of trick would allow you to run some 200B parameter models on the same hardware - the scale where AI models stop being stupid and useless.
  
  
  
• And if you have an Acemagic F9A, 1.6T parameters.  (WCCFTech)
  
  You'd want a fast SSD for that, though, because that would be a lot of data to fling about.
  
  
  

Musical Interlude

• So how are things going a year into the DRAM Apocalypse?  (Tom's Hardware)
  
  I'm glad you asked.
  
  Actually, no I'm not.
  
  But anyway, you remember when DDR5 memory was brand new and people preferred DDR4 because it was much cheaper?
  
  Yeah.  DDR5 costs twice as much now as it did back then when it was too expensive to be worthwhile.
  
  
  

Tech News

• Speaking of DDR4, how do Intel CPUs fare when using it?  (Tom's Hardware)
  
  Okay-ish.  The tricky part is when you see benchmark numbers for older AMD chips like the 5800X3D you know they are using DDR4, because that's all they support.  Intel's 12th, 13th, and 14th generation chips all support both DDR4 and DDR5, and using DDR4 can lop off 14% from published benchmarks.
  
  
  
• If you send Claud Opus the character sequence - - - it writes a novella about an unnamed architect.  (Austin's Nerdy Things)
  
  Sadly only via the API.  I tried it in the web interface and it just said "Huh?"
  
  
  
• AMD's new Radeon 9050 graphics card is half a 9060.  (WCCTech)
  
  It comes in 8GB and 4GB flavours, each with half the compute cores of the 9060, and the 4GB model having half the memory bandwidth as well.
  
  
  

Posted by: viralnudez at Saturday, August 08 2026 12:27 AM (CCnh7)

Hide Comments | Add Comment