Thank you for your sacrifice in tracking this down.
I only take the time to discuss this pattern as a way of showcasing how information travels on Twitter in the modal case
In the same genre: my attempt to map out the early-2025 freakout about some alleged insane breakthrough at OpenAI.
I suppose one annoying counter-example to this pattern is September 2026 rumors about a frontier lab resolving a Millennium Problem, which were also warped, but ultimately did originate from something true (the Levent/Tristan collaboration leaking). That unfortunately means that "Twitter noise is always bullshit" is not quite the right model to have... But you should probably internalize, at the gut level, that it ought to be the default interpretation (instead of "there's no smoke without a fire"), and you should only dig into it if you really have nothing better to do.
It certainly wouldn't change anything if I took the time to reply to every one of these examples with my lowbie account.
That's what Community Notes is for right?
Or how about "@grok is there actual concrete evidence to substantiate this rumor"
Even if the information on twitter is often false, I'd still vastly prefer having one place on the internet where we can have a lot of false information, than having traditional media be the entire arbiter of truth. Without twitter, we'd be stuck in the lies and propaganda of traditional media entirely.
In this case. I believe it's directionally correct to model what happened as some type of power struggle in the government between factions that were okay with letting jensen do what he wanted to do, and factions that weren't. And that jensen got ousted as he basically lost this power struggle. This is just normal politics.
So although the specifics of what twitter is saying in this case are probably incorrect, we don't have to accept the official story. I believe it's already not clear if she was really a spy for example? Wasn't that the official story? Or is it an 'alleged' spy? Whatever either way.
I think the main reason to worry about “traditional” media is the possibility of a biased and censored central authority. Yet, I think state level actors can still exert huge influence on decentralized media - pushing stories they want to see with bots or sock puppets, under the cover of “free journalism.” Let alone the power of who sets the algorithm.
Fair enough.
In this case. I believe it's directionally correct to model what happened as some type of power struggle in the government between factions that were okay with letting jensen do what he wanted to do, and factions that weren't. And that jensen got ousted as he basically lost this power struggle. This is just normal politics.
I don’t think it’s “directionally correct” because the conspiracy theory you’re describing is completely unsubstantiated.
It's really, really bad if you call things completely unsubstantiated conspiracy theories and they turn out to be potentially correct. Like that's how people get polarized against traditional media (see: lab leak, WMD, etc)
It's pretty clear that there was a faction (led by Rubio and Bessent) who wanted tighter chip controls, vs Lutnick and Sacks who sided with Jensen.
It seems like the initial leak probably came from someone in that former faction.
Am I missing anything here?
CNBC: Bessent blames Lutnick’s Commerce as he slams Nvidia chip ‘leakage’ to China
CNN: Intel intercepts implicating Jensen Huang in affair sparked classified Rubio memo
I am also confused by the above comment. My post literally makes the same suggestion, and cites a Washington Post article about it. I think people are just not reading my post.
Disappointed by the extreme response to OP, whose comment is totally reasonable. @Garrett Baker @Dean Valentine @warner @uugr @Viliam what are you supposing is the source of these alternative perspectives that would exist in the non-Twitter hypothetical, such that OP's claim would be wrong?
The main claim that Jensen's firing was the result of a power struggle within the admin is also totally true and confirmed by mainstream sources like the Washington Post (a source told the Post that the underlying memo and leak aimed to "destroy the influence Jensen had over the president"); the Cabinet was divided between a pro-export control faction that included Rubio, Bessent, and Hegseth and an anti-control faction that included Lutnick, Sacks, and Huang. The "spy" claim has also been walked back by official sources like the attorney general, the FBI director, and the media. The Singapore Home Affairs ministry also published the questions and response from the WSJ in order to reject the Journal's claims that Singapore was investigating Alice Huang for espionage.
Without twitter bluesky would exist, so would mastodon, reddit, substack, lesswrong, facebook, discord, slack, and many other websites which often don't simply parrot or are actively skeptical toward mainstream reporting.
The claim that "Without twitter, we'd be stuck in the lies and propaganda of traditional media entirely." Is simply absurd on its face, though there may exist weakenings which are less obviously absurd, we then need to contend with Valentine and others' arguments against Twitter, and may then conclude that one should use alternatives to Twitter which have less of a mind-melting, shit-eating dynamic to them.
Bluesky is a funny example because while Bluesky would exist if Twitter were destroyed today, Bluesky only exists because of Twitter's current direction. Facebook is not the same type of open platform, nor is Discord or Slack. Reddit has much more intense moderation and policing of non-mainstream views.
LessWrong is good but doesn't cover the wide range of political topics; for example on Twitter, accurate info can be found about news stories related to race, sex, immigration, or free speech itself, such as the ICE/CBP shootings, the Lucy Craig case at Cornell, "two-tier policing" in the UK, criminal prosecution of Michel Houellebecq, or whatever; LessWrong doesn't treat this topics and is overall a smaller platform.
A totally unique place in the media environment. It's the only place where you are more likely to see inadequately challenged right-wing misinformation or misleading content than inadequately challenged left-wing misinformation or misleading content. That makes its marginal value extremely useful for gaining an accurate perspective. As an example I see tweets from this journalist Michael Tracey who seems to mainly get attention by poking holes in the stories of alleged rape victims, including in the Cornell case but also Epstein island and other cases. I have no strong opinion and have done no deep research on my all-things-considered views on these cases, but his perspective is uniquely valuable because it is hard to find elsewhere in the same simple presentation.
Can you point to cases where Facebook, Mastodon, or whatever provides these perspectives?
Michael Tracey also has a Substack where he seems to write about the same stuff, and Substack in general has a large conservative element to it including names like Curtis Yarvin and Richard Hanania. There is also the podcasting ecosystem which also has a large conservative following including the following podcasts (listed by Claude because I'm less familiar with this ecosystem):
Claude's conservative podcasts
The mass-audience tier
Edison's ranking puts Ben Shapiro, Dan Bongino and Tucker Carlson at the top for conservative content.
The ones that covered Arjun's exact cases
Higher-quality, lower-heat
This is the tier I'd cite to a LessWronger, since "inadequately challenged misinformation" is less of a worry here.
Right-adjacent or heterodox
These aren't conservative, but they carry the same contrarian-on-sex-and-race perspectives.
Sources:
I'm not on Facebook so idk any specifics there, and Mastodon/Bluesky are indeed very liberal, but I would argue not in a "traditional media liking" way. Similarly, though Reddit is generally liberal and generally traditional media liking, it is very big and very diverse so there are many parts of reddit which are not so traditional media liking, such as The Motte (which I haven't visited for a while, and it turns out they moved to a new standalone website, though they do still post their quality contribution threads to the old reddit. Notably they didn't move to twitter).
I would also call your claim here a weakening of the original claim, and also not entirely absurd. Twitter may indeed currently be a great place to find very right-wing pushback on mainstream media (I think this is a broadly reasonable opinion to have). That is different from the claim that without it we would "be stuck in the lies and propaganda of traditional media entirely" for two reasons:
Good point about Substack. I don't use it that much, and don't use its Substack Notes feature.
I think your podcast examples are weak; I don't know much about them, but my impression is that they're not very informed or else just kind of like lowest-common-denominator slop.
More generally I feel like
I upvoted the OP but disagree voted.
I didn't downvote OP (just disagree-voted and -reacted), but I don't agree that Twitter is providing a unique or valuable service by collecting lots of low-quality conspiratorial information in one place. There are lots of other places one can find large collections of low-quality conspiratorial information (Garrett Baker lists some in his comments). But even if Twitter were the only place poised to offer such, I'm not sure why this is desirable. Twitter is full of rumors and conspiracies, most of them are wrong, and there's no clean way to separate the wheat from the chaff (or the incorrect from the incorrect-but-kind-of-shaped-like-something-other-people-later-found-out-was-true) purely from within the website.
The "directional correctness" claim sounds like it's suggesting that a Twitter user, though misled by the rumor mill, would at least be misled in a way that puts them closer to what actually happened than an MSM reader. To me that seems like a bad heuristic. Reversed stupidity is not intelligence; you don't get better information by listening to people who are wrong in the opposite direction to the mainstream. You might be occasionally vindicated when one of the rumors is true, but not in a way that generalizes to forming better beliefs overall.
Is there a pattern of thought you think would correctly deduce the real power struggle from looking at the Twitter rumor mill, but wouldn't over-update on the false details (eg that the anti-Huang faction was masterminded by Jack Clark), and would be able to do the same for other rumors as well?
I'm confused as to what your position is.
The people you mentioned disagreed with this line from HowManyAgents:
Without twitter, we'd be stuck in the lies and propaganda of traditional media entirely.
You say:
[people you disagree with above line], what are you supposing is the source of these alternative perspectives that would exist in the non-Twitter hypothetical, such that OP's claim would be wrong?
You then go on to say:
The main claim that Jensen's firing was the result of a power struggle within the admin is also totally true and confirmed by mainstream sources like the Washington Post (a source told the Post that the underlying memo and leak aimed to "destroy the influence Jensen had over the president”)
Are you not answering your own question? The alternative perspective is the Washington Post. You are yourself citing the "lies and propaganda of traditional media" (which I wouldn't characterize as such and may well be why five people signaled disagreement).
I cited the WashPo post in order to convince/appeal to those who trust it. The post was published the day after Jensen resigned i.e. after the damage was mostly done. Twitter users were aware of the weaknesses in the "spy" characterization and in the obvious motives of the IC sources for the e.g. NYT post as early as the day it came out, which was the actual key period for the public and elites to reacts, which affects the likelihood of resignation or board action etc.
Did you notice that my post already cites the same article:

I did not read your post (until I skimmed it just now). I only saw the downvoted comment and response to which I replied.
Where did you come across that passage from WashPo?
No I mean, how did you learn the passage existed (or that you were interested in reading the article as a whole, if you found the passage by reading the article)
Background research for my blog.
I appreciate you taking the time to do this because I would never choose to do so.
One attached thought is cognitive science work relating to cultural transmission which studies how the human centipede process tends to converge to the priors of the members of the centipede: https://onlinelibrary.wiley.com/doi/10.1080/15326900701326576.
In this case though the rate of transmission isn't only controlled by the likelihood the members of the chain assign to a particular idea like we would expect in offline communication but that bayesian process plus the effects of the algorithmic reproduction process. In general I think these kinds of chains often tell us more about the priorities of the company running the platform than the actual beliefs of the population.
X/Twitter rewards content that generates engagement. It also doesn't have, in my opinion, any effective mechanism for penalising inaccurate statements[1]. Community notes and corrections in the comments do exist, of course, but in expectation they're seen by far fewer people than the original post (e.g. the "allegedly" comment).
An analogy I like to use to illustrate the fundamental issue is how research papers shouldn't cite Wikipedia. The information may well be substantially correct; the problem is that citing a tertiary source that anyone can edit at any time is, structurally, a weak epistemic foundation. X is even weaker, since there's no easily verifiable chain of provenance or attribution[2].
My personal rule is to treat X as a good source for first-hand statements from named people, and completely disregard everything else.
Also, I suspect people are significantly more comfortable being confidently wrong or making claims they haven't verified on X than they are in person. Penalising mechanisms such as "feeling shame" seem to be significantly weaker in online environments. Caveat: I'm not a psychologist and this is "pop psychology", hence the footnote.
I acknowledge "traditional" media has its issues too. Still, I think on balance the odds of being misled on X, compared to a news organisation with an established reputation for factually accurate reporting, are meaningfully higher.
I used to think Community Notes by X was enough to counteract this sort of rumour mill, but I guess it's not enough? I'm not sure what's required, but it feels like something more powerful than per-tweet annotations is necessary here.
The comment seems to imply that the solution is finding a new tool for patching X rather than considering that a better solution is credible companies, researchers, engineers and nonprofits taking their expertise and conversation elsewhere.
I think figuring out a technological solution to the problem would in fact be revolutionary for humanity's collective epistemics, yeah. And it's not obvious to me that one couldn't exist.
For clarity, so would I. My point is more about how we are framing the problem. We can be in agreement, for example, that building safe airplanes is extremely important, while also raising concerns about the people actively sabotaging their engines before takeoff. ’We should focus on building safe airplanes’, while true in isolation, then seems to miss a crucial part of the problem context.
On October 29th, 2026, the New York Times reported that Jensen Huang had a ‘close relationship’ with Alice Huang, an executive U.S. officials had concluded was a Chinese spy. The story had six reporters on the byline, and it was sourced to "people briefed on the surveillance."
Whether or not you think Alice Huang was working for the MSS, a rational person might conclude that the person "briefed on the surveillance", who supplied the initial story, was a U.S. government official. After all, if you were not possessing a security clearance and working for U.S. government, you would not be getting "briefed" on U.S. counterintelligence investigations.
As the story gained traction this month, people who think quite differently began to propose other theories. On November 10th - the first time I can find this suggested - an account called "ellington" stated their "personal conspiracy theory" that Anthropic was the New York Times' source (2k likes):
A few hours later GeLun Ding, a Chinese-language account that shows up again below, said the same thing without hedging (1.4k likes):
On November 11th, 6PM Pacific Time, Dream1984Tube, another Chinese-language account announced that he had "found the key clue" to who is really behind the hunt for Jensen Huang, and that he was writing it up:
Twenty minutes later, he says he needs to take another look at Ana Swanson, one of the six Times reporters. At 6:32 PM he brings up Jack Clark (cofounder of Anthropic) for the first time, on the grounds that Clark went to an Aspen Institute dinner with one of the six reporters:
On November 11th at 9:14 PM PT, he posted an AI-generated longform article on X titled "Who Is Really Behind the Hunt for Jensen Huang?" This is the first post, on or off twitter, that accuses Jack Clark of being the source behind the October 29th article, and it's the first one that takes off:
Most of the content is devoted to describing second- and third- degree connections between the people who talked about or expressed indignation at the affair. It contains only the following core points:
At this point in Twitter history, this is the entirety of the concrete evidence offered, that Jack Clark was the source for the October 29th New York Times article. His name has been repeated many times on X throughout the saga. As far as I can tell, no one has ever attempted to explain how Jack Clark would know who Jensen Huang is sleeping with.
Grok's post is then amplified by other mouthbreathers in the tech industry, much more than the previous tweets:
And within the hour, "Hensen Juang", a South Asian twitter account, performs the standard magick whereby inferences are transmuted into observations.
He first posts a couple "suggestive" tweets:
And then, in his most popular tweet, he restates the main insinuation from the article without qualification, and receives thousands of likes:
Pulled from Twitscan, which shows UTC (9:52 PM PT)
An hour and a half after posting, he adds a one-word reply consisting of the word "Allegedly", which gets an order of magnitude less circulation than the original tweet.
And a day or two later after that he deletes the original tweet, after about a million people have already seen it:
The rumor continues to spread, in various formats, mostly across other Chinese language accounts (screenshots cropped to fit on the page, so that you can see tweet metrics):
Simultaneously, a game of telephone begins.
During a podcast, Tripp Mickle discusses how the New York Times first started looking into MegaSpeed last year, for an article published on October 9th, 2025 titled 'Is China Buying Banned Nvidia Chips?':
To re-emphasize some relevant context here:
Context notwithstanding, here is how Philippe Lemoine reports this interaction on X The Everything App, at 6am on November 12th:
Philippe's tweet heavily implies that "the story" that "didn't originate with the intelligence community" is the one that kickstarted Jensen's firing. A casual reader would incorrectly assume that a "tech industry contact" is who told the New York Times about Jensen's affair, which is naturally what Twitter users do.
The next day, Chief Nerd, an account with nearly a million followers, cites the same clip but this time claims that the "Jensen Huang story" was the one tipped off by the "tech industry contact:"
And about this time is when Chinese language accounts tweet the same link, transmuting "a tech industry contact" into "Anthropic", in order to get an order of magnitude more uptake:
This post, by "Dragonhau66", is what is finally screenshotted and converted back into English by other Chinese-speaking users. During the translation, "Anthropic" is transmuted further into specifically "Jack Clark, co founder of Anthropic":
And about then is when the "story" really takes off:
Behind show more: "And what do you know, the person who leaked the “honeypot” story to the press was none other than ANTHROPIC CO-FOUNDER Jack Clark, who reportedly holds a larger stake in Anthropic than Dario Amodei himself…"
Etc., etc.
It's probably worth noting that back in the land of legacy media, Marco Rubio was already named as the author of the memo warning Trump about the relationship between Huang and a suspected Chinese spy. According to the Washington Post:
The Post article describes the memo's authorship as part of a power struggle within the executive branch, over whether Nvidia should be allowed to sell chips to China. I don't think Rubio leaked the memo, but if people feel like being conspiratorial, it seems much more plausible that he circulated the memo knowing there was a chance it would be leaked.
It is now Sunday and the story is still ongoing; I saw more and more people casually mentioning the "Anthropic coup" while doing research to write this post. I suspect I will continue to see references to Jack Clark's treachery for months to come. Doing this trace took way more time and energy for me than it did for these individuals to write their tweets in the first place, and my trace will not spread nearly as far. It certainly wouldn't change anything if I took the time to reply to every one of these examples with my lowbie account.
I don't really care about this particular topic, or about Jack Clark. I only take the time to discuss this pattern as a way of showcasing how information travels on Twitter in the modal case, and to gesture at the hundreds of other well-circulated bits of false information I've observed, which no one ever bothers to "debunk" in some visible way. If you use Twitter to keep up with AI, especially if you tell yourself that you're "protecting" against the malicious influence of traditional media, you too will fall for commensurately more plausible examples of this pattern. No one is going to take the time to correct your misconceptions.