Walter Appling retweeted
Even with Jensen Restorationists in Silicon Valley, I'd figure the NVIDIA board may want to keep their distance from this criminal investigation which clearly has NVIDIA and Huang at the center of it.
Exclusive: After the Biden administration restricted sales of Nvidia's A100 chips to China in 2022, Nvidia sales employees suggested Alibaba rent the chips from Alice Huang's company, Megaspeed, in Southeast Asia.
The U.S. Justice Department is now leading a criminal investigation of Megaspeed.
Full story: thein.fo/4kHscrF
Walter Appling retweeted
The only people on planet earth who believe that are the ever-dwindling band of misfit MAGA cultists.
Walter Appling retweeted
Trump is a gormless oaf with zero awareness he’s an imbecile. Xi and Putin, however,are fully aware that’s he’s an easily manipulated gormless oaf who will give them the cow in exchange for three magic beans.
Walter Appling retweeted
Donald Trump and Howard Lutnick were co-conspirators with Jensen Huang in the largest technology smuggling operation in American history.
Well well well, Jeffrey Kessler, who runs the BIS, rushed staff to approve Nvidia H200 exports to China; some chips were granted to companies tied to Chinese military or smuggling networks
big scoop by @QianerLiu @leomschwartz via @theinformation $NVDA
theinformation.com/articles/…
The Trump administration approved shipments of nearly 900,000 advanced AI chips made by Nvidia and AMD for Chinese companies from December 2025 through May 2026, the House Foreign Affairs Committee said reut.rs/4AKntdn
Walter Appling retweeted
Republicans attack GTA 6, 'It sets a bad example for children when innocent people can be violently abducted from their cars by non-government employees'
Walter Appling retweeted
Every GOP nominee who waffles on questions about January 6 or the 2020 election should get treated exactly like this.
Sen. Thom Tillis (R-NC) has reportedly placed a hold on the nomination of Michael LoSapio after the nominee stated that the Capitol was not attacked on Jan. 6, 2021.
Trump tapped LoSapio, a former executive at Palantir, to be director of the Cybersecurity & Infrastructure Security Agency at the Department of Homeland Security
mediaite.com/politics/republ…
Walter Appling retweeted
This is so cruel to the police victims of the US Capitol siege and their families.
This madness is likely to end on Jan 3, 2027.
Hard to imagine nominees will be confirmed by a Democratic majority in the US Senate if they continue to answer like this:
(Via MSNOW)
Walter Appling retweeted
Just so you fully get what you are seeing today--both Russia and China have been helping Iran attack US soldiers and US friends.
Trump is literally celebrating with people who have aided the killing of American service personnel.
Walter Appling retweeted
"Measles is mild."
Unless you get something called SSPE
It causes kids to forget how to speak, walk and swallow.
And then they die.
SSPE has no cure. It's always fatal.
But measles vaccine prevents it .
Walter Appling retweeted
One might even say you got "deported" from your country club...
Walter Appling retweeted
Trump: I love the farmers. I’ve done more for farmers than any other president!
Walter Appling retweeted
The leader of the Republican Party is allied with Putin.
There is nobody in the Republican Party who will do anything about. Very few have the courage to even complain about it publicly.
Which means the Republican Party is allied with Putin.
It is the Putin Party.
Walter Appling retweeted
These past two years, it became frustratingly clear that a whole bunch of people who claim to hate corruption meant they hate corruption in the other party, and don’t get all that upset when the accusations and allegations involve members of their preferred party. A lot of people on the right who were justifiably spitting mad about the deep-pocketed folks throwing money in various forms to Hunter Biden had much less to say about a Russian oligarch paying for Donald Trump Jr.’s wedding party. (Meanwhile, here at NR, Noah wrote about it, Dan wrote about it, I wrote about it, we talked about it on The Editors.)
Walter Appling retweeted
Blah blah. 1) There can be no "NATO-like" security guarantees as long as Trump is president because he will never act against "his friend" Putin and Putin knows it.
2) Russia agreeing to something is meaningless because it is a terrorist mafia state that freely violates any agreement when it suits.
3) Giving up Ukrainian territory and people to reward Russia's aggression guarantees more aggression and provides advantages for it.
4) Like all of these sham negotiations, their goal is to pressure Ukraine into surrendering while delaying action by Europe and the US Congress and normalizing Russia to get sanctions lifted--money that goes into the war effort.
5) Putin cannot end the war; it's necessary for his survival in Russia. Even if he were given everything he demands, aggression in Europe would soon continue, and escalate.
6) The more success Ukraine has on the battlefield and with long-range sanctions against energy and infrastructure, the more Putin will push propaganda in the West and for these charade summits and negotiations.
Walter Appling retweeted
Yes, Oren. When you torture data to try to extract the opposite of what it shows, and when you trumpet the product as a headline result even though it's undermined by your full dataset, it's statistical fraud.
And that fact is not obviated by footnoting your torture techniques.
"Statistical fraud" is a pretty serious accusation, especially when the purported "fraud" is an intentional methodological choice clearly stated and explained in the report itself, and more so when the purported effects of the "fraud" do not exist.
A tweet is not the place to explain what a regression is, but I'm sufficiently embarrassed for some of the people I see blindly retweeting this that I do think it's worth the time to set the record straight.
The critique that dropping voters with no issue variance leaves a more pro-R sample and thus will leave tariffs looking more popular is a category error. If the analysis were just adding up the number of people who said each issue moved them one direction or the other, that would be a problem. But that’s not what we’re doing.
Or, more specifically, that is what we do as a preliminary step, and in taking that preliminary step we do show the raw sums for the entire sample, because that's the right way to look at a raw tabulation. It's the first chart on this analysis in the report, helpfully titled "Voters reported D push everywhere."
But this is not the question we really want to answer. What we want to know is: can we find a signal in the way strength of views across various issues actually correlates to vote cast. Now we're not summing across all responses, we're looking at the relationship between responses for each voter. In doing so, looking at responses where there is no variance across views is pretty unhelpful.
More than half of respondents gave the same answer on all five issues. Whatever moved them, their answers can’t tell us which issues mattered more. So, as we make very clear in the report, we drop voters whose push across issues did not vary. Note that, once we've dropped them, we never show another raw tabulation again, like the initial one, which included the full sample. Because that would be biased and misleading. But the correct, raw, unbiased view is already there.
OK, now, what ARE we doing with this smaller sample? We’re running various forms of statistical regression, which ask how different responses on different issues correlate to actual votes. A regression does not sum up the preferences of the people in the sample, it looks at the relationship between responses for each person in the sample. In short, it asks of each issue: holding views on all other issues constant, how much does the Democratic vote share shift when views on a particular issue shift? You really want to know this for people who differentiate among the issues. There is almost no useful signal about which issues mattered in the responses of someone who says all issues mattered the same.
This becomes painfully obvious if we just add all the voters back in. This is an easy robustness check. And guess what? It doesn’t change the conclusions at all (though immigration looks more beneficial to Republicans). This makes sense if you think about how the statistical models work. The larger "unskewed" sample actually reduces the tariff effect. If the goal were to minimize tariffs in the results, using the full sample does that best. We didn’t do it that way because we don’t think it’s the right model. Table with full results below.
One more point: when we bring our logit model back and apply it to the electorate to see the full vote swing, we DO apply it to the full sample again. We thought about this one a lot, because you could definitely make the case that if the model is built on a subset of voters then it should only be applied to the subset of voters.
But that WOULD be subject to the critique that we were modeling a skewed sample. And it would assume an electorate in which everyone has variable responses to issues, which obviously is not right. So we applied our estimates to the full sample and then provided a robustness check: "Applying the logistic model to only the 847 respondents who differentiated among issues yields larger effects for corruption and the Iran war, but tariffs remain statistically insignificant."
I appreciate Phil. He does more to undermine market fundamentalism than anyone else on the Internet. But it makes me sad to see people like @scottlincicome push this stuff around and celebrate the substitution of half-assed AI slop for the harder work of thinking through the issues.
I can see why it would be frustrating to spend months tweeting PMIs, just to have the PMIs turn positive, then tweet about employment, until employment turned positive, then... but there really is greater dignity to be found in acknowledging mistakes and working constructively toward better policy than in just broadcasting propaganda and rooting for your own country to fail.
When you see an analysis you disagree with, start by asking the person what they did and why. Did they consider the alternatives? Have they checked for robustness? What do they think of your approach and would they be able to test that as well? That’s how you make progress in debates and on policy. This is something much uglier.
Walter Appling retweeted
He has a child’s view of many things. He’s basically an 80-year-old toddler.





































