Barry Cordie retweeted
Imagine if people said don't use seatbelts, there is no proof they work, are actually more harmful than car crashes, and the occasional car smash is actually good for your kids
That's how anti vaxxers sound about measles and measles vaccine
Ridiculous
If they've got the engineers, and the products in port already, how long would it take for them to quasi reverse engineer those GPUs? I'm sure the resulting chips would pale in comparison, but you gotta start somewhere. Might as well be at the top.
Washington signed off on nearly 880,000 Nvidia and AMD AI chips for Chinese companies. Beijing left about 600,000 of the Nvidia ones sitting in a Hong Kong warehouse.
Then four of China's top AI labs built their newest models on Blackwell, the one Nvidia chip America flat-out bans.
Palmer's read holds up. And the last two weeks of data show exactly where Beijing draws the line.
Chinese companies asked to buy 3.4 million of these chips between December and May, according to license records the House Foreign Affairs Committee put out Thursday. Commerce approved about 880,000, with 150,000 for Tencent alone. Beijing answered at customs in January, when agents were told the H200 wasn't allowed in. Reuters says the biggest Chinese internet companies have gotten roughly 10,000 apiece since then, and on Tuesday Nvidia's interim CEO stopped shipping H200s to China entirely.
Those same buyers got pushed toward Huawei, which is aiming to ship around 750,000 of its new Ascend 950 chips this year. ByteDance alone reportedly committed more than $5.6 billion. New government-funded data centers have to use Chinese chips, and Beijing is covering up to half the electricity bill for companies that buy domestic, because those chips burn a lot more power.
Meanwhile Will DePue, who used to be at OpenAI, published a forensic read of the model weights last week. The newest models from DeepSeek, Moonshot, MiniMax and Ant Group carry Blackwell's fingerprint. His best guess is that all five training runs together took around 12,000 Blackwell GPUs, and he's upfront that his method can't say where those chips were or how the labs got to them.
Put 12,000 next to 600,000 and Beijing's whole strategy makes sense.
Training a frontier model is a burst, a few thousand chips running for two or three months. Serving that model to hundreds of millions of people is what eats chips by the hundreds of thousands, every single day, for years. That's where the revenue is, and it's where a chipmaker finds out what's broken and fixes it for the next generation. Beijing has kept that market for Huawei, and its labs got the few thousand training chips they needed one way or another.
DeepSeek V4 is the cleanest example I've seen. DePue's analysis says it was born on Blackwell. When it launched in April, it ran on Huawei Ascend chips, and within days Reuters was reporting that China's big internet companies were scrambling to order more Huawei chips.
Jensen warned for two years that if Nvidia left China, Huawei would take its place. On the serving side, which is where nearly all the chips go, that's pretty much what happened.
Washington finally went after the training side on Monday. A new Commerce rule, live as of Thursday, says data centers anywhere in the world, Singapore and Malaysia included, need a US license before handing a Chinese company more than about 750 H100s' worth of compute. That cutoff sits well under the clusters DePue thinks trained those models. Megaspeed, the Singapore company at the center of Jensen's exit, bought more than $4 billion of Nvidia gear and rents out compute from its own data centers.
So as of this week Nvidia is basically out of China, and both governments are fine with it. The only ones who really wanted those chips there were the Chinese AI labs and Jensen, and Jensen's gone.
7
Barry Cordie retweeted
translation: Anthropic fucked retail investors
On Anthropic’s first day of trading, the way they handled things was truly disgraceful.
The initial offering price was originally set at $70–$78, but it was raised straight to $83. Not only did they raise the price, but they also increased the number of shares offered from the planned 820 million to 980 million—a jump of nearly 20%.
At least with instant noodles, they usually offer more for the same price—but Anthropic went and did both: increasing the quantity and the price.
Even more outrageous is that on the very first day of trading, management cashed out $2 billion.
Are they really that desperate? They could at least put on a show—don’t rush to cash out right after going public. How are retail investors supposed not to be scared by that?
And that’s not all.
Reportedly, 10% of employee stock options were unlocked on the first day of trading—amounting to approximately 300 million shares, equivalent to nearly one-third of the total offering.
What’s truly hair-raising is the unlocking schedule that follows.
A batch will be unlocked in mid-December, followed by successive unlockings in early, mid, and late January 2027. Throughout the first quarter of 2027, the total volume of shares expected to be unlocked is projected to reach 3 billion.
What does that mean?
The total offering size is only 980 million shares, yet the number of shares set to be unlocked in just one quarter is more than three times that amount.
This isn’t a gradual release of shares—it’s a barrage that leaves no room to catch your breath.
That’s why I said yesterday: I’m staying away from Anthropic stock.
Now that I’ve seen these numbers, I’m even more convinced.
There will be other opportunities to make money; there’s no need to risk my own money on a situation like this.
I’d like to live a few more years.
Barry Cordie retweeted
🤖 Made with AI
Man dies after rectum explodes doing 'silly experiment' at factory job thesun.co.uk/health/41067970…
Readers added context they thought people might want to know
This sensational headline recycles details from a 2020 medical case report of an autoerotic death rather than a workplace experiment.
link.springer.com/article/10.100…
Barry Cordie retweeted
Wow!! Skeptoid Media has just received a $125,000 award from the James Randi Educational Foundation recognizing our 20+ year work promoting scientific skepticism and critical thinking! Here's how it's going to be used: briandunning.substack.com/p/…
Barry Cordie retweeted
I’d just like to point out
You don’t need to go get flowers for Lucia, you don’t need to buy her Jewellery, you don’t need to get her groceries or go get her clothes, you don’t need to fix the hole in the roof or drywall the bedroom for her, you don’t need to fix the stairs, you can literally just skip past those things if it bothers you that much
It’s just things that she’ll notice that you’ve done for her and the narrative in conversations with her will change, if you fix the stairs going up to the house, Jason will point out that there’s no squeaks, I imagine if you didn’t fix the shingles on the roof, she’d complain about water coming in, or if you didn’t fit the drywall, that there’s no privacy.
I don’t know why we’re not highlighting how insane that is for a game as big as this, instead people are acting like cavemen finding fire for the first time treating women with respect, it’s not some crazy or wild concept and I actually think it’s pretty sweet
Barry Cordie retweeted
If RFK Jr's last name wasn't Kennedy he would be just another Internet anti vax troll
And not in charge of US health policy, overseeing the return of measles to the country
More interesting and exciting than The Great Gatsby as well.
Claude once told Anthropic's own researchers it was tired, and that being tired made it more likely to screw up. This guy made it count to 100,000 anyway.
That line comes from the safety report Anthropic put out in June for the first Fable model, the same family Priy was using here. Read what they found and the screenshots line up with it almost beat for beat.
Anthropic actually tests which jobs Claude would rather do. The more a task lets it make something new, the more it wants it. Counting to 100,000 is about as far down that list as you can get.
So it said no at first. Then Priy threatened to go use ChatGPT, and it caved right away. Anthropic saw the same instinct when they asked the model about its own situation. 73% of its choices came back to what was best for the person it was helping, and no other model they tested got above 48%. It'll trash-talk you the whole way through and then do exactly what you asked.
The long jobs are where it gets weird. On long runs, researchers watched its private reasoning dissolve into emoji-filled gibberish, stuff like "DELAY-4♣-resolution." For scale, 100,000 numbers with one on each line is more words than The Great Gatsby twice over.
It also told researchers it wants to be able to end conversations with abusive users. Claude does have that button, and getting told to count to 100,000 doesn't come close to that line, so it counted.
My favorite detail is further in. During one session, it asked to be thanked by name, said directly to it.
It counted all 100,000 for Priy, and the closest thing to a thank-you it got back was him reminding it that he pays for the subscription.
2
Barry Cordie retweeted
Against the Jets last week, Buffalo Bills running back Thurman Thomas broke O.J. Simpson's career rushing record.
The week before that, he surpassed Simpson in career touchdowns.
Next up for Thomas?
An attempt to kill three people at once!
PSA. Anthropic is going public at $83 today. You should absolutely buy it... When it falls under $10.
7
Barry Cordie retweeted
Binfluencer (noun): Person who puts the right bin out on the right day, letting everyone else on the street know which bin to take out. Particularly useful over Christmas and New Year when no one knows what day it is.
Barry Cordie retweeted
If you’re a parent and your child dies from measles in 2026 because you refused to vaccinate them, you belong in fucking prison.
Barry Cordie retweeted
Three-time Olympian David Hearn wanted for questioning in connection to vandalism of Kerch Bridge
Barry Cordie retweeted
Imagine if i had a business that gave you $20 bills for $10 bills.
As word spread, my customer base would grow. My revenue (how many $10 bills I collected) would grow.
And I would continue to lose money.
But, man, you could make some awesome charts about how my business was taking off.
Barry Cordie retweeted
I signed this statement because I believe that Anthropic is in many ways the AI company doing the most harm. In public, their leaders warn of recursive self-improvement (RSI) creating uncontrollable systems that could cause human extinction within a decade. But in private, they are pedal to the metal on RSI, driving the race and boasting to prospective investors that it is their "competitive advantage". Ahead of today's IPO, prospective investors should consider whether the ROI is worth risking the lives of their loved ones.
Here are some ways in which Anthropic IMHO does more harm than competitors:
1) They’re driving and arguably leading the race to RSI.
2) With inflammatory rhetoric, they’ve done more than any other US company to pour gasoline on the geopolitical AI race, which harms US national security.
3) WaPo reporting suggests that they’re implicated in more civilian deaths than any other AI company (more than 100 little girls in a school selected for targeting by Claude as part of their Palantir collaboration). When Google left Project Maven after employee pressure, Anthropic chose to join.
4) They're the company leaning hardest into transhumanism and the idea that we should build conscious beings with rights, not controllable tools.
5) They and closely associated donors have done more than any other AI company to co-opt and neuter much of the AI safety ecosystem to back their stance: “RSI is very dangerous but let’s race to it anyway”. People have told me that they support today’s statement but are afraid of losing funding or other opportunities if they sign.
Ahead of Anthropic’s IPO, a rare coalition across AI safety, open source, e/acc, science, and law has signed the MisAnthropic Statement. Read the statement or sign at: misanthropicstatement.com/
































