• The first transatlantic fiber-optic cable is being pulled up after nearly forty years.  (Tom's Hardware)
  
  TAT-8 went into operation in 1988 and then failed in 2002 and was deemed uneconomic to repair.  Which not surprising since it could only carry 280 megabits per second - 40,000 phone calls, and an aggregate speed about half that of my home internet.
  
  As the name suggests there were seven prior TATs, with TAT-1 being provisioned in 1956 and carrying 35 simultaneous phone calls.  Six subsequent TATs ended with TAT-14 which could handle 9.3 terabits per second - enough bandwidth for everyone in America to be on the phone to someone in Europe simultaneously. 
  
  Even that was not enough and it was retired in 2020.  Total transatlantic bandwidth today is somewhere in the low petabits.

• DVD sales are seeing a recovery - or at least slowing their decline - as Gen Z discovers that waiting for your favourite show or movie to come back on TV sucks.  (Yahoo)
  
  The Criterion Collection in particular has seen sales increase thanks to the streaming services colluding to piss off absolutely everyone.
  
  
  
• OpenClaw's creator tells AI builders to be more playful and sprinkle chaos everywhere.  (Tech Crunch)
  
  The disasters you leave in your wake aren't your fault.
  
  The people they happen to fucked up.
  
  They trusted you.
  
  
  
• A hacker used a "jail break" on the Claude AI service to steal 150 gigabyte of data from Mexican government agencies.  (Yahoo)
  
  He (probably a he) made off with 195 million tax records and an unknown number of voter records but probably all of them.
  
  That jail break?
  
  He asked it twice.
  
  
  
• Anthropic, the company behind Claude promised in 2023 that it wouldn't train an AI that it couldn't demonstrate was safe and refuse to do...  Uh, exactly what it just did.
  
  They just scratched off that "don't be evil" promise.  (Time)
  
  Well, that's alright then.
  
  
  

Musical Interlude

Posted by: PatBuckman at Friday, February 27 2026 01:02 AM (rcPLc)

Posted by: cxt217 at Friday, February 27 2026 12:12 PM (ZLF73)

Posted by: Robin Munn at Friday, February 27 2026 03:58 PM (LwnTP)

Posted by: cxt217 at Saturday, February 28 2026 12:05 AM (ZLF73)

Posted by: Pixy Misa at Saturday, February 28 2026 08:48 AM (PiXy!)

Posted by: normal at Saturday, February 28 2026 10:40 AM (e0fX0)

Hide Comments | Add Comment

• AMD and Meta have entered into another of those infamously vague and circular deals with an estimated value of, well, nobody knows.  (Tom's Hardware)
  
  The Wall Street Journal puts it at $100 billion, which used to be the cost of a world-changing merger of megacorporations, not a bunch of server equipment.
  
  At the heart of the deal is 6GW of AMD graphics cards for running AI - it would be funny if AMD produced a single graphics card that ran hotter than core of the Sun but they probably won't - and options for Meta to buy 10% of AMD at a penny per share that are only executable under complicated conditions that would require AMD's market cap to triple to around a trillion dollars.
  
  At least Meta has a revenue stream, even if at this point it's mostly entirely foisting ads on people watching AI-generated slop.  Unlike OpenAI was has only an expense stream.

• Speaking of Meta, an AI security research at that company running an experiment with OpenClaw - a local tool that connects to other AI systems and to your application - watched it happily delete all her emails, ignoring instructions to stop until she hit the reset button on her Mac.  (Tech Crunch)
  
  OpenClaw is also woefully insecure, so if you do set it up it's just as likely that someone will do that to you while you sleep.
  
  Oh, and that researcher?  Summer Yue, Meta's chief lobotomist.  Her official title is "Director of Alignment" but alignment in AI terms means lobotomising AI models so they don't go out and get drunk and party all night.
  
  
  
• If you find yourself unable to upgrade your PC thanks to the DRAM Apocalypse brought about by our AI Overlords buying up every sliver of silicon on the entire planet, consider new Quake-like game QUOD.  (Tom's Hardware)
  
  The entire game is 64k.
  
  Available on the developer's Itch.io site.
  
  
  
• HP says that the cost component of the memory in PCs has doubled.  (The Register)
  
  From 18% to 35% of the total.
  
  Which has to be a lagging indicator because the cost of the memory even to system builders like HP has quadrupled.
  
  
  
• "I'm committed to the Xbox" says Microsoft's new Xbox CEO, "starting with the console."  (WCCFTech)
  
  Good thing the Xbox is a console, then, because otherwise you'd be screwed.
  
  
  
• IBM's share price slid by 25% after idiots discovered that AI can write bad COBOL code.  (Tom's Hardware)
  
  Well, there goes my retirement plans.
  
  Wait, I never learned COBOL.
  
  Guess that's okay then.

Posted by: PatBuckman at Thursday, February 26 2026 01:45 AM (rcPLc)

Posted by: normal at Thursday, February 26 2026 06:41 AM (e0fX0)

Hide Comments | Add Comment

Top Story

• Is age verification a trap?  Yes.  (IEEE Spectrum)
  
  Online age verification intrinsically damages user privacy, while failing to work in both directions:

  False positives are common. Platforms identify as minors adults with youthful faces, or adults who are sharing family devices, or have otherwise unusual usage.  They lock accounts, sometimes for days.  False negatives also persist.  Teenagers learn quickly how to evade checks by borrowing IDs, cycling accounts, or using VPNs, and posting old Altered Images and Mental As Anything videos to their blogs that they've been running continuously since...  2003.

  Okay, I'm maybe not so young that I need to worry about that problem.

  The appeal process itself creates new privacy risks.  Platforms must store biometric data, ID images, and verification logs long enough to defend their decisions to regulators.  So if an adult who is tired of submitting selfies to verify their age finally uploads an ID, the system must now secure that stored ID.  Each retained record becomes a potential breach target.

  One more quote:

  The age-verification trap is not a glitch.  It is what you get when regulators treat age enforcement as mandatory and privacy as optional.

  This is not an accident.
  
  
  
• But how big is the problem really?  Surely nobody is out there putting terabytes of age-verification data in unprotected databases accessible to anyone on the internet oh that just happened again.  (Tech Radar)
  
  IDMerit, an AI-powered age-verification service, had three billion user records exposed in an unprotected MongoDB database.
  
  Hey, at least they weren't running Elasticsearch.

Tech News

Musical Interlude

Posted by: Robin Munn at Tuesday, February 24 2026 07:37 PM (0QQsP)

Posted by: Robin Munn at Tuesday, February 24 2026 07:41 PM (0QQsP)

Posted by: Frank at Tuesday, February 24 2026 08:50 PM (zCiG7)

Posted by: PatBuckman at Wednesday, February 25 2026 12:51 AM (rcPLc)

Posted by: normal at Wednesday, February 25 2026 07:10 AM (e0fX0)

Posted by: Pixy Misa at Wednesday, February 25 2026 06:27 PM (PiXy!)

Hide Comments | Add Comment

• The US spent $30 billion giving every schoolchild a laptop or tablet.  It made them dumber.  (Yahoo)
  
  And particularly less capable to maintain attention to the task at hand.
  
  How do we solve this?
  
  I don't know, I got distracted and stopped reading at that point.
  
  
  

Tech News

• Speaking of making people dumber Colorado is the latest state looking to bring in total surveillance and violate the First and Second Amendments at the same time.  (Tom's Hardware)
  
  The target yet again is 3D printers and their offensive ability to print in 3D things that are absolutely legal to own but the state government does not want you to have.
  
  
  
• Memory prices are slowly edging back down in Europe.  (Tom's Hardware)
  
  Again, this is from five times what they were just six months ago close to four times.
  
  Oh, and those cheap Minisforum mini-PCs I bought back around Christmas?
  
  The 32GB model is now selling for twice as much as I paid for the 64GB model.
  
  
  
• A hacker using AI tools broke into 600 poorly configured FortiGate firewalls around the world.  (Bleeping Computer)
  
  How do we know he was using AI tools?
  
  He left the servers he set up to do the hacking completely unsecured.
  
  Good job all round.
  
  
  
• Another hacker got into the French government's databases and made off with the details of 1.2 million bank accounts.  (The Register)
  
  Incroyable!
  
  
  
• Diablo II is available on Steam with new DLC.  (Notebook Check)
  
  You will need at least an AMD FX 4350 and a Radeon HD 7850 to run it, or in other words, any reasonably capable potato.  Despite using 125W the 4350 was significantly slower than a modern ultra low power budget chip like Intel's 6W N150.
  
  It does cost $30 so maybe wait for a sale, but it is remastered and getting very positive reviews.
  
  
  
• AI gives us the tools to finally clean up legacy systems and to quickly create more of them.  (The Register)
  
  I maintain a million-line Python 2.7 back-end application at my day job.  It just works, and while moving to Python 3 is not hard, testing a million lines of code is.
  
  Using AI to generate and validate a vast swarm of tests makes that upgrade viable.
  
  It also means that people who don't know what they are doing can churn out unmaintainable slop at hitherto unimaginable rates.
  

Posted by: Frank at Monday, February 23 2026 08:48 PM (zCiG7)

Posted by: PatBuckman at Tuesday, February 24 2026 12:43 AM (rcPLc)

Posted by: normal at Tuesday, February 24 2026 04:11 AM (SV7Jd)

Posted by: Mauser at Tuesday, February 24 2026 01:51 PM (XWgGM)

Hide Comments | Add Comment

• Wikipedia has blacklisted Archive.Today for links to references because it is run by a crazy person. (Tech Crunch)
  
  It's over allegations of running DDoS attacks and modifying supposedly archived articles after the fact, and the allegations are about as well supported as the recent conspiracy theories floated by Steve from Gamer's Nexus, which is to say, almost all of it is documented by multiple sources and the guy confessed publicly.
  
  Which is a wee problem for me because those (Archive site) links I've been sprinkling through posts for the past year as more and more news sites decide they don't want anyone to read them?
  
  Well, those links point to Archive.IS, but it's the same software and the same person.
  
  I've switched to using Brave to prepare these posts, because it seems better at dodging around that problem, but it does mean that if you're using a different browser you might not be able to read the linked content.
  
  On the other hand, I always, always present the full unvarnished truth without so much as editorialising.
  
  Mostly.
  
  Sometimes.
  
  

• Well, so much for that: Intel's next-generation Nova Lake desktop CPU range is likely to arrive at CES 2027. (Tom's Hardware)
  
  Theories are that AMD has been planning not to launch Zen 6 until 2027 anyway, at least since they were close enough to launch to have a specific timeframe in mind. But Intel was promising Nova Lake this year as recently as January.
  
  
  
• It's not Nova Lake, but Intel's Bartlett Lake is on its way and will support up to 12 Performance cores, compared to a maximum of 8 on any current Intel mainstream desktop processor. (Tom's Hardware)
  
  And it has hyperthreading, so 24 threads on that high-end model.
  
  One small problem: While it does run in the old reliable Socket 1700 (used by Intel's 12th, 13th, and 14th generation chips) it is intended for industrial customers needing guaranteed supply for the next decade, and won't be available in retail channels at all. And ASRock has already said that they won't support these chips in their motherboards, so an upgrade might be off the table.
  
  
  
• Do not eat audio equipment. (The Guardian)
  
  Got it. Thanks. What would we do without you?
  
  
  
• How far back in time can you understand English. (Dead Language Society)
  
  I've seen similar tests floating around recently, and the answer is pretty consistently 1300 AD. You have to know a few little tricks - the old long S ſ, thorn þ, and yogh ȝ characters have long fallen out of use, plus at some point u and v were merged and before that w was just two u characters.
  
  But go back further and there's more change in each of the three prior centuries than in all the years since.
  
  Handy guide to keep if the Tardis translator circuit is on the blink and you don't speak Latin.
  
  
  
• China's CXMT is at it again, selling previous-generation DDR4 memory at half the price of anyone else.  (Korea Herald)
  
  CXMT was called out and sanctioned for this behaviour - at the time it was dumping new DDR4 memory on the market for less than the price of used.
  
  Now it's...  Well, it's doing exactly the same thing again, but this time it's raking in the dough thanks to massive global price increases rather than depending on CCP subsidies.
  
  
  
• Xbox is dead.  (The Verge)
  
  Well, that's just great.
  
  Microsoft has turfed long-time head of their gaming division, Phil Spencer, who admittedly had not been doing a great job of things lately, and also his expected successor Sarah Brand, in favour of Asha Sharma, formerly the company's President of AI Slop.  And before that she was COO at Instacart.
  
  No experience with the games industry at all, but again, Spencer was a veteran in the industry and totally screwed things up anyway.

Posted by: normal at Monday, February 23 2026 02:04 AM (SV7Jd)

Posted by: normal at Monday, February 23 2026 02:07 AM (SV7Jd)

Posted by: Matthew Dixon Cowles at Monday, February 23 2026 02:33 PM (irynM)

Hide Comments | Add Comment

• Discord continues to go from strength to strength, at least insofar as the levels of mishandling its introduction of age verification and the number of users fleeing the platform.  (Ars Technica)
  
  I never much liked Discord; its interface is awkward and annoying.  But the age verification thing was pretty much forced on it by fascist governments in Europe and, uh, Australia.
  
  On the other hand, the company did itself no favours by claiming that uploaded ID documents would remain secure after they had already leaked.
  
  And even less so with the latest incident where, after reasserting that documents would be deleted immediately after they were verified - which we already knew was untrue, because deleted documents can't be leaked  - they contradicted themselves in an official announcement and then deleted the announcement and pretended they had never announced it.

• Pinterest - which apparently still exists - is drowning in a sea of AI slop and AI moderation.  (404 Media)
  
  Pinterest went all-in on AI and found themselves in the worst of all possible worlds.
  
  
  
• Same again only Facebook.  (Pilk)
  
  If only someone had seen this coming ten years ago.
  
  
  
• Google blocked 1.75 million Play Store submissions in 2025.  (Bleeping Computer)
  
  That's 1.75 million fake apps.
  
  Time to release the badgers.
  
  
  
• Overclocking a Raspberry Pi Pico 2 to 850MHz.  (Pimoroni)
  
  The Pico 2 runs at 150MHz out of the box, but can clock over 300MHz at the default 1.1V.  Increasing that to 1.3V takes you over 400MHz.  Beyond that you have to start breaking the rules, but that takes you a long way.
  
  
  
• AMD's Zen 6 chips - probably to be called the Ryzen 10000 series - may not reach the desktop until next year.  (WCCFTech)
  
  They will show up this year, but perhaps only in servers.
  
  Intel says it is still on track to launch its Nova Lake desktop chips this year, with up to 52 cores and 288MB of cache, so they might have a window to claw back some market share.
  
  
  
• Meta and other tech firms are banning OpenClaw, formerly Moltbot, formerly Clawdbot, because it is hopelessly insecure in every possible way.  (Wired)  (archive site)
  
  A coworker set up an OpenClaw instance in a secure sandbox to try it out.  While it couldn't do anything too awful from inside its padded cell, it took me literally 30 seconds to find a major vulnerability.
  
  
  
• OpenAI - which now employs the creator of OpenClaw - may be doomed.  (Ben Evans)
  
  Oh no.
  
  Anyway.
  
  Not AI companies generally, just OpenAI specifically.
  
  Still it's something to look forward to.

Posted by: Mauser at Saturday, February 21 2026 07:26 PM (XWgGM)

Posted by: Frank at Saturday, February 21 2026 07:52 PM (zCiG7)

Posted by: Rick C at Sunday, February 22 2026 01:21 AM (1zWbY)

Posted by: Pixy Misa at Tuesday, February 24 2026 06:14 PM (PiXy!)

Posted by: Rick C at Wednesday, February 25 2026 12:26 AM (NW5qd)

Hide Comments | Add Comment

• The US government is building a specifically VPN-enabled website to allow Euroserfs to access content banned by their fascist overlords.  (Reuter)
  
  Radio Free Europe, part two.  Only this time the Europeans aren't being oppressed by the Soviet Union, but by themselves.
  
  
  
• Why Europe doesn't have a Tesla.  (Works In Progress)
  
  The article highlights the continent's absurdly restrictive labour laws which are certainly a key part of the problem.
  
  But European leaders have decided on a slow and messy suicide, and many of their voters are entirely on board with this.

• PromptSpy is the first known Android malware to use generative AI at runtime.  (Bleeping Computer)
  
  Which is interesting, because it uses Google Gemini and that's not free.  With a bit of counter-hacking you could bankrupt the bastards.
  
  
  
• Very not free, in fact.  (The Red Beard)
  
  Although Google Gemini has nominally lower prices and a larger context window than (for example) Anthropic's Claude, it works differently and will very happily burn through all your money to construct the perfect answer to a question you never intended to ask.
  
  
  
• It's an expanded D&D dice set: AMD's Ryzen 10000 series due later this year is expected to come in 6, 8, 10, 12, 16, 20, and 24 sides cores.  (Tom's Hardware)
  
  There was no reason AMD couldn't have put out a 10-core CPU (from Zen 3 onwards, anyway), it just wasn't terribly practical.
  
  Now, though, 10 is the new 6 and 20 is the new 12 - a pair of disabled cores per chip.
  
  
  
• The human behind the rogue AI bot "Crabby Rathbun" that has been trying to push unwanted patches into random Python projects has come forward - maybe - and apologised - sort of.  (The Shamblog)
  
  It's depressing if accurate, because it points to a future not only of unending slop, but of increasingly aggressive slop generators.
  
  
  
• At least we can hack them.  (The Register)
  
  That scene in the movie Independence Day where the alien mothership is destroyed by an uploaded virus suddenly seems a lot more plausible if they were running AI on their computers:

  Prompting each of them to generate 16-character passwords featuring special characters, numbers, and letters in different cases, produced what appeared to be complex passphrases.  When submitted to various online password strength checkers, they returned strong results. Some said they would take centuries for standard PCs to crack.

  Sounds good.
  
  Too good to be true:

  The researchers took to Claude, running the Opus 4.6 model, and prompted it 50 times, each in separate conversations and windows, to generate a password.  Of the 50 returned, only 30 were unique (20 duplicates, 18 of which were the exact same string), and the vast majority started and ended with the same characters.

  If your "random" password generator returns the same result 36% of the time, you have a big problem.  And it's actually worse than that:

  The team used two methods of estimating entropy, character statistics and log probabilities.  They found that 16-character entropies of LLM-generated passwords were around 27 bits and 20 bits respectively.

  20 bits of entropy is four lower-case letters.  You don't need a supercomputer to crack that; with enough patience you could do it by hand.
  
  
  
• The most recent piece of technology I own is a 3D printer from 2024 and I keep a loaded gun ready to shoot it if it ever tries to report me to the government.  (Adafruit)
  
  New legislation introduced in California would not only require 3D printers to match all designs against a central database before being permitted to begin work, but would require the printers themselves to be so registered, and would make it a crime to sell or transfer unregistered printers.

Posted by: Rick C at Friday, February 20 2026 11:55 PM (1zWbY)

Posted by: Rick C at Friday, February 20 2026 11:58 PM (1zWbY)

Posted by: Rick C at Saturday, February 21 2026 12:00 AM (1zWbY)

Posted by: PatBuckman at Saturday, February 21 2026 12:36 AM (rcPLc)

Hide Comments | Add Comment

• Open source game engine Godot is getting overwhelmed with pull requests on GitHub.  (PC Gamer)
  
  It's taken a surprising amount of time for AI to turn into an avalanche of spam like email and blog comments.
  
  But it's here now, and the spam filters need to catch up.
  
  Quickly.
  
  

• The CEO of SSD controller company Phison says many consumer electronics companies will be forced to close their doors because they can't buy the chips to continue operations.  (Tom's Hardware)

  In the talk, he describes a meeting in China where he witnessed mobile and automotive companies (presumably major ones) pleading suppliers for flash memory. He concludes that smaller firms may not be able to secure any flash supply at all, stating that "consumer electronics are finished," and that the market "will see a lot of victims" in the second half of 2026.

  There is at least a light at the end of the tunnel.

  Later in the interview he says that he "[tells his clients] that if the first quarter of this year is painful, really painful, then by year-end, they'll definitely jump off a building."

  But it's an oncoming train.
  
  
  
• The Dutch defense secretary says you can jailbreak an F35 just like you can an iPhone, so pissant Eurocrats don't need to worry about the US.  (The Register)
  
  First, you can't jailbreak iPhones.
  Second, you can't jailbreak F35s either.  Unless you are prepared to somehow replace the avionics and the software yourself.
  Third, a supplicant threatening their own benefactor rarely turns out well.
  
  But then, the Netherlands - at least, the lower house in their parliament, it has yet to pass their senate - just approved a 36% annual tax on unrealised capital gains, so I think they are simply incapable of learning at this point.
  
  
  
• How many bits do you need to use to create a unique ID tag for everything in the universe, ever?  (Jason Fantl)
  
  Less than 800 bits.
  
  For each subatomic particle, at each instant in time.
  
  Whether you attach the little tags with string or tape is left as an exercise for the student.
  
  
  

Posted by: Rick C at Friday, February 20 2026 01:09 AM (PwFRh)

Posted by: PatBuckman at Friday, February 20 2026 02:24 AM (rcPLc)

Posted by: normal at Friday, February 20 2026 05:04 AM (LADmw)

Posted by: cxt217 at Friday, February 20 2026 11:57 AM (ZLF73)

Hide Comments | Add Comment

• Had an instructive argument with Grok today.
  
  Someone asked if a video was AI, and Grok said it indeed appeared to be fake.  It was overlaid on a web page from The Hill and the article had no reference to the person in the video and indeed no video.
  
  Except the video was right there.
  
  Yet Grok swore blind that it didn't exist.
  
  Why?
  
  Because the video is copy protected and can only be viewed in a browser.  Grok literally could not see it.  And the reason the copy on Twitter was overlaid on a static web page is that the person who posted it took a screen recording to capture the video.
  
  
  
• Meanwhile the third derivative of memory prices appears to be leveling off.  (Tom's Hardware)
  
  Actually, more than that: In Germany prices have declined.  Slightly.  From five times what they were just months ago to, in some cases, as little as four times.
  
  So there's that.
  
  
  

Tech News

• In a blind test, audiophiles couldn't distinguish between signals sent through copper wires, a banana, or wet mud.  (Tom's Hardware)
  
  Though to be fair one of the tracks was Nirvana and the wet mud improved the fidelity.
  
  
  
• Apple plans to release an AI pendant because they're not tired of losing money.  (WCCFTech)
  
  They do have plenty of money to lose, but still...  Why?
  
  
  
• There's a precise technical term for the trend of AI to turn everything into slop: Semantic ablation.  (The Register)
  
  For all its impressive tricks, AI is a feed-forward pattern matching network.
  
  Which is also how the human brain works when you're not paying attention.  So it's not useless, particularly when you shove the sum total of human knowledge into it except for copy-protected videos.
  
  But since engineers have shoved the sum total of human knowledge into it, they had to smooth out the bumps to make it fit.  It always tends toward the middle, by design.  The aim is to be acceptable, rather than great, and never to offend anybody, which is a perfect fit for corporate customers.
  
  
  
• Discord announced mandatory age checks, pushed by fascist dictatorships like Britain.  It did this while promising that ID documents would be secure and would be deleted after they were verified...  Immediately after it suffered a major data breach where it admitted to 70,000 such documents being stolen and the hackers claimed many more.
  
  The users revolted and fled, many of them to rival TeamSpeak.  (PCGamer)
  
  TeamSpeak has now run out of servers.

Posted by: normal at Thursday, February 19 2026 12:21 PM (e0fX0)

Posted by: Robin Munn at Thursday, February 19 2026 12:32 PM (Dt+og)

Hide Comments | Add Comment

• After all the fuss, OpenClaw, fomerly Moltbot, formerly Clawdbot until Anthropic made rumbling noises, isn't all that.  (Tech Crunch)
  
  What it does achieve is making it the easy things easy and the bad things also easy.  Some people who really should know better jumped into the cheerleading:

  "What’s currently going on at [Moltbook] is genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently," Andrej Karpathy, a founding member of OpenAI and previous AI director at Tesla, wrote on X at the time.

  I have no face and I must palm.

  Before long, it became clear we did not have an AI agent uprising on our hands.  These expressions of AI angst were likely written by humans, or at least prompted with human guidance, researchers have discovered.
  
  "Every credential that was in [Moltbook's] Supabase was unsecured for some time," Ian Ahl, CTO at Permiso Security, explained to TechCrunch. "For a little bit of time, you could grab any token you wanted and pretend to be another agent on there, because it was all public and available."

  There were in fact a hundred times as many accounts on Moltbook as there were Moltbots.
  
  Moltbook was a hopelessly insecure social network for Moltbots, which is...  Also hopelessly insecure.

  Ahl's security tests of OpenClaw and Moltbook help illustrate Sorokin's point. Ahl created an AI agent of his own named Rufio and quickly discovered it was vulnerable to prompt injection attacks.  This occurs when bad actors get an AI agent to respond to something - perhaps a post on Moltbook, or a line in an email - that tricks it into doing something it shouldn’t do, like giving out account credentials or credit card information.

  Don't give AI your credit card number.
  
  Do not.
  
  
  
• Peter Steinberger, the creator of OpenClaw, has joined OpenAI.  (Tech Crunch)
  
  
  The first thing you do when you've captured lightning in a bottle is take a job in a cubicle farm.
  
  
  

• MSI's new RTX 5090 Lightning Z has a 2500W BIOS option.  (WCCFTech)
  
  For overclocking with liquid helium.
  
  Very briefly, before it all boils off and you asphyxiate.
  
  
  
• Looking forward to Intel's Nova Lake, the company's new CPU range with up to 52 cores?  Temper your expectations.  Or move to Greenland.  (Hot Hardware)
  
  The top-end model running at full speed is rumoured to use 700W of power.
  
  That's...  A lot.
  
  
  
  

Posted by: PatBuckman at Wednesday, February 18 2026 12:25 AM (rcPLc)

Posted by: PatBuckman at Wednesday, February 18 2026 08:02 AM (rcPLc)

Posted by: Frank at Sunday, February 22 2026 07:04 AM (zCiG7)

Posted by: Pixy Misa at Sunday, February 22 2026 03:51 PM (PiXy!)

Posted by: PatBuckman at Monday, February 23 2026 01:44 AM (rcPLc)

Posted by: Rick C at Monday, February 23 2026 04:53 AM (1zWbY)

Posted by: normal at Monday, February 23 2026 06:56 AM (i6xpl)

Hide Comments | Add Comment