I don't think you understand the objection that I and I assume many others have had to him. His objections are very non-random/motivated and this only becomes apparent after repeated interactions with him. I like having scrupulous friends, and even more I like having scrupulous friends who aren't too burned out from addressing non central objections to actually discuss the ideas that they want to discuss. Scrupulous people are nerd snipeable. So I clear the area of snipers when I can.
His objections are very non-random/motivated
I don't think that matters, because the implied alternative of offering only "random", "unmotivated" objections (a) is not humanly possible and (b) may not even be well-defined. It's true that any given commenter has their own "beats", themes, and pet peeves, because they're individual humans and not the all-knowing God-Empress. That's why it's important to get comments from lots of different people!
addressing non central objections
The problem with this is that people trying to protect their ideas from scrutiny have an incentive to reframe objections as "non-central" in a way that doesn't actually hold up, such that we want the job of determining whether it's central or not to belong to the readers rather than the author. If I make a claim and offer some evidence, and someone points out that my evidence doesn't support the claim, it's tempting for me say, "Oh, well, you're just nitpicking, that's non-central, my claim is still true" even though I wouldn't have agreed the evidence was "non-central" if no one had challenged it.
Scrupulous people are nerd snipeable. So I clear the area of snipers when I can.
I think either your friends are not actually scrupulous, or you're being a bad friend. As a scrupulous person, I want to hear the non-central objections! I care about whether all the things I say are true or not, not just the "central" things! (I might not get around to replying, but that's OK: readers can read the comments for themselves and decide for themselves.) Managing my own effort and burnout risk is my job; I don't need to censor other people's speech to do it! Right? What is the scrupulous reason to disagree with this?
As a grown-up on an intellectual discussion forum, it's not other people's job to manage your feelings.
It absolutely is.
Of course, the trick here is the fuzzy meaning of "manage your feelings". It's something that both parties have to do. Yes, people must moderate their own feelings and not act emotional about every little thing. But likewise, it is the responsibility of the poster to not attack others and justify it by claiming that it's just feelings and it's not the poster's fault if someone gets upset from being attacked.
It does not make sense to object that the intellectually substantive comment makes you feel bad, because whether the comment makes you feel bad has no bearing on whether the comment is true or relevant.
This is the autistic quokka attitude: autistic because of the assertion that feelings don't matter and quokka because you need to recognize attacks.
If the intellectually substantive comment makes you feel bad because of its intellectual content, for instance, because you don't like being proven wrong, you might have a case. If it makes you feel bad because of other reasons, then no. It's not at all hard to write something which intermixes intellectual content and personal attacks, and we need to be able to ban those, not say "reacting to attacks is feelings. Nobody should be concerned about feelings, and the intellectual content is still sound".
By your reasoning if he was in your house tossing live grenades and making arguments, it would be bad to kick him out of your house, because the grenade explosions don't affect the logical consistency of his arguments.
Zack: As a grown-up on an intellectual discussion forum, it's not other people's job to manage your feelings.
Jiro: It absolutely is.
I'm with Zack on the quoted sentence. Typically when a person doesn't like something, that because there's something bad about it [the thing they didn't like]. It's generally good to avoid bad things. But... I think we get far healthier patterns when we make it peoples responsibility to avoid causing particular kinds of bad things, than when we make it peoples' responsibility to manage other peoples' feelings (even if done by both parties).
In terms of things (that seem to me to be) near your (Jiro's) statements that I can agree with:
I’m not going to express any further opinions about Said here but agree otherwise with this post; it seems to me that the explanation of the ban was not an honestly mistaken application of good principles, but an appeal to bad principles. If an appeal to feelings can be reduced to legitimate considerations motivating those feelings then one should do so; if not, the appeal is illegitimate. I took the ban as a confirmation of a policy of unprincipled and unbounded submission to threats by the people (or personas) whose feelings are supposed to matter, which seems to be disjoint from those who mean to be accountable for their speech and behavior.
it seems like they're trying to strike a balance between ego gratification and being less wrong. You can get away with making on-topic, intellectually substantive criticisms, but only if you take care to manage ingroup members' emotions and cover for their reputations
I think this is core to the question. It's been a main point of discussion since the inception of LessWrong and the split from Overcoming Bias - how to balance the needs of anchor posters and the rest of us. Ego gratification is a big draw, and it likely keeps the site healthy overall to notice and prevent particularly grating responses and demands.
EVEN IF those annoying-to-ingroup behaviors are legit in a lot of lenses, they're harmful to the feeling of community. I don't honestly know if it's possible to have a feeling of community without in-group enforcement of norms that aren't ideal.
I appreciate your work on many issues, but find myself pretty strongly disagreeing.
finding minor criticism of a sub point that is technically valid (often only if you interpret things noncentrally) doesn't mean the broader claim is wrong.
Said imposed more cost then provided value, and refused to cooperate with extensive efforts to find third options. Banning that seems productive.
I. Prologue
"If I Can't Explain It to Said Achmiz, I Probably Don't Understand It"
This post isn't really about him, but I'd like to begin with a tribute to my friend Said Achmiz, the wisest person I know.
The choice of adjective is deliberately chosen as term of art. Achmiz is not the most quick-witted, nor the most knowledgable, nor the most creative, nor the most savvy. I say wise because the nature of his peculiar talent is the same one ascribed to Socrates: he knows what he doesn't know.
Most people don't. At least, I don't. All too often, when asked, "Do you get what I'm saying?", I reply, "I think so," with perfect sincerity, only realizing later that I was lying—that I could not have paraphrased my interlocutor's ideas, let alone explained where I agreed or disagreed with them. In retrospect (and only in retrospect), it's clear that I didn't want to disrupt the harmonious give-and-take flow of social exchange, that I feared being seen as foolish or obstinate for having the temerity to say, "No, I don't understand"—and to continue standing by that No should understanding not be forthcoming.
The unique genius of Achmiz is simply that he's not afraid to be seen as foolish or obstinate. If he doesn't understand how you're using a term, he'll ask for a definition. If he doesn't see how to apply your abstract claim, he'll ask for an example. If your offered definition or example doesn't make sense to him, he'll say so. Bluffing doesn't work on him.
Occasionally it strikes me as a little odd when Said asks about some detail that I think he should have been able to fill in by himself, but it's no trouble: if I can spare the time (and it's worth it), I'm usually happy to type up the requested additional explanation. And if I can spare the time but can't come up with an explanation that will satisfy him, that's an alarming warning sign that my grasp of the topic wasn't as firm as I thought: I didn't know what I didn't know.
A few years ago, I was disappointed when the conclusion of a post that I had labored over for months didn't land in the comment section the way I wanted it to: what my words had meant in my head, wasn't the same as what the commenters took them to mean.
To help me figure out how to revise it, I emailed Said, explaining what I meant. He had some criticisms. I proposed a rewrite. He said the rewrite was still confusing. I wrote some additional explanation, particularly of an analogy in the passage in question which I was quite fond of. Said was unmoved and pointed out some relevant disanalogies. I cut the analogy from the rewrite entirely, which met with Said's tentative approval.
The revised passage is less engaging writing than the original. (It was a really clever analogy.) But it's clear, and for that I have Said to thank—not just for his generosity with his time, but for his steadfastness. A lot of people in that situation wouldn't have stuck to their position over such a long back-and-forth (and back-and-forth-and-back-and-forth): their acute social perception would have noticed that I was fond of the analogy and "mercifully" told me what I wanted to hear—at the cost of making the post worse. [1] In contrast, Said, no less socially perceptive, has the strength of will to not make sociality his master. The plea of a friend and the threat of an enemy are equally ineffective at budging him from calling it as he sees it.
I've been blessed with many friends and collaborators in this life, each with their own unique skillset. Some of them are prolifically generative, gushing with a dozen theories and metaphors and frames at the slightest prompt. Some of them know extremely advanced mathematics and can prove in minutes what would take me hours just to understand the theorem statement. Some of them are kind and compassionate, ready to lend a sympathetic ear to anyone when the world's cruelties seem too much to bear.
Said is none of these things. But when what I need is a critic's discrimination—when I'm worried that I might be fooling myself via lack of rigor—I can't think of anyone I'd rather have by my side. If I can't explain it to Said Achmiz, I probably don't understand it.
II. The Ban Contradicts Less Wrong's Stated Values
An Inversion of Values
Less Wrong chief moderator Oliver Habryka begins what I can only describe as an astonishing document by contrasting two frequently voiced perspectives on the culture of Less Wrong.
On the one hand, it is said that "LessWrong is a place that really forces you to get your arguments together", that its straight-shooting culture is "one of the things that makes it one of the most valuable forums on the internet."
On the other hand, it is said that using the website "seems really unrewarding" because:
Habryka opines that "both of these perspectives are right."
I would go further. It's not just that both of these perspectives are right. It's that you can't have one without the other. By what means can a web forum "really force you to get your arguments together"?
There is only one means: by commenters pointing out all the places where a post says something wrong—anything wrong. If the error wasn't "load-bearing for your argument", it's still worth pointing out in the comment section, for the benefit of readers who might otherwise be deceived. (We're not going to run out of paper.) It doesn't matter if it "isn't something you consider particularly important." If it's so unimportant that you don't even want to receive corrections, then what was it doing in the post?
If the "discussion doesn't illuminate what you are trying to communicate," it's worth asking: why is that? If it were because the commenters were for some perverse reason pretending en masse to not understand your clear and correct explanations, then that would be bad. But if it's because your explanations are not, in fact, clear and correct, then disappointment was inevitable: commenters can't illuminate what you're trying to communicate if you didn't successfully communicate it.
As a perennial Less Wrong author of over 100 posts—including some nine Curated posts, including four voted Best of Less Wrong in 2019 (#9), 2021 (#31), 2023 (#14), and 2024 (#33)—I have the experience to testify on this matter. Yes, it is disappointing when a post doesn't land in the comment section the way I wanted it to, but it's obviously not because commenters are culpably withholding their illumination of my precious ideas. It's because writing about ideas is hard and I don't always get it right. Sometimes I'm wrong, sometimes I'm confused, sometimes I don't write clearly—and it shows in the comment section and the karma score. I certainly prefer it when the discussion illuminates what I was trying to communicate, but the only real way to accomplish that is by me writing something that commenters as individuals find illuminating—and there's always the possibility of that not happening when people are being honest, rather than colluding to flatter the author's ego. "Yes, and—" requires the possibility of "No, because—".
(As for "implying that they are superior in their dismissal of your irrational and dumb ideas", I suppose it's true that when someone rejects an idea, that creates a logical implication that the rejecter thinks they're more rational than the idea-proposer on that topic. It's a weirdly petty implication to focus on, though. Who cares?)
What is most striking about the second perspective's list of reasons that critical commenters make Less Wrong unrewarding to interact with is what it does not say.
It doesn't say the commenters are wrong! If the second perspective had claimed that commenters mischaracterize posts (rather than merely criticizing non-"load-bearing" aspects of the posts or failing to "illuminate what [authors] are trying to communicate"), that would be a serious problem that would warrant corrective actions by moderators. We can infer that bad arguments and strawmanning are not seen by user reports as a pervasive problem on Less Wrong. (If they were, the second perspective would have mentioned it.)
Thus, it would seem that the two complementary perspectives described by Habryka are describing a success story of a vibrant intellectual forum with a culture of excellence that cares fanatically about getting things right via the power of vigorous and rigorous discussion. It's true that posting on Less Wrong can be "unrewarding"—but that's because the rewards are real and therefore have to be earned.
Habryka states that "few people have done as much to shape the culture of LessWrong" than Said Achmiz, and in the comments guesses that a survey would place Achmiz around #20 on a list of such figures (and that "this would be an underestimate of [Achmiz's] effect on the culture").
At this point, some inattentive readers who missed the title of Habryka's post might be under the impression that Achmiz is being honored with some kind of award. (Why else would Habryka be writing a post describing what's good about the website's culture, and then naming Achmiz as a cornerstone of that culture? What else would the post be announcing?)
Upon realizing that the post is announcing a ban of Achmiz, such a reader is at a loss at how to react: the decision would seem to represent a complete inversion of the stated values that purportedly justify it. The fact that Habryka has written a 9000-word post explaining his reasoning offers hope that some clarification will be provided, but upon reading it, the mystery only deepens. It's hard to even know where to begin. The intimidating task of dissecting the post's errors and omissions is one that would rightly demand the talents of a discriminating critic of the highest order, someone like ...
Well, I guess that's my job now.
If "Every Individual Owns Their Own Judgment of You", Then the Ban Rationale Is Untenable
Habryka goes on to articulate two alleged failure modes of online communities. In the sneer attractor, people tear others down with nonspecific criticism; in the LinkedIn attractor, people build each other up with nonspecific praise. Habryka goes on to say that many users see Achmiz as serving a vital function of preventing Less Wrong from falling into the LinkedIn attractor. He identifies Achmiz as the "bearer of a flag" which is worth quoting in full:
Well spoken! Habryka reports that he "really care[s] about this flag too" and that "much of the decisions [he has] made around LessWrong have been to foster a culture that understands and rallies behind this flag." He expresses hope that his acknowledgement of the risks of LinkedIn-like sycophancy will give supporters of the culture of excellence "a bit more trust that [he is] tracking some things they care about."
However, it is difficult to reconcile this claim with the remainder of the post, which persistently construes intellectually substantive criticism as "punishments" or as claims to have "violated some social norm", and ultimately announces a ban of Achmiz on the grounds that the judgments Achmiz expresses have deleterious social effects (bringing the site closer to the alleged sneer attractor). [2] The entire point of the post is to announce that the moderators are preventing Achmiz from openly discussing Less Wrong authors and their contributions!
After a discussion of Achmiz's alleged commenting patterns, Habryka speculates that users will "ultimately give up feeling dejected and like a lot of people on LessWrong hate you." Setting aside the obvious irrationality of this inference—there's no good reason to think that one user's persistent questioning of your ideas implies that "a lot" of people "hate" you—it's hard to see the relevance given Habryka's supposed fostering of a culture that understands and rallies behind the aforementioned flag. Habryka just told us 900 words earlier that just because you're anxious about others criticizing you or your ideas doesn't mean the website is going to accommodate you! Anyone can scroll up and read it. And yet when people are hurt by and anxious about Said Achmiz in particular criticizing their ideas, the response is to ban Achmiz? The disconnect is glaring.
The Ban Reneges on LessWrong 2.0's Single Locus of Discussion Mandate and Archipelago Doctrine
In the ban announcement, Habryka writes that "the foundation of why [he] feel[s] comfortable governing LessWrong with relatively few checks and balances" is because people have the option of just not using the site: "[t]here are many other places on the internet to read interesting ideas, to discuss with others, to participate in a community," he writes.
In contrast, in September 2017, Habryka's "LW 2.0 Strategic Overview" cited Anna Salamon's "On the Importance of Less Wrong, or Another Single Conversational Locus" as something "very important to [him]." Habryka quotes Salamon (bolding mine):
That is, in Salamon's telling, apparently endorsed by Habryka in 2017, the reason to reboot Less Wrong was not just to be a place on the internet to read interesting ideas, discuss with others, or participate in a community, but specifically to reverse the "diaspora" that occurred when the original implementation of the site fell into disrepair—to be the place to read interesting ideas building off the legacy of the Sequences.
Reversing a diaspora only works if the diverse elements of the diaspora are allowed to participate in the repatriation. For example, the State of Israel's legitimacy as a homeland for the Jewish people depends on its including all Jews as the concept has been historically and publicly understood. If a faction of Ashkenazi Jews seized power and arbitrarily decided that Sephardic Jews can't be Israeli, that would seriously undermine the foundations of why many people supported the Zionist project in the first place. [3]
Said Achmiz and his collaborators (such as the present author) are undisputedly part of the rationalist tradition. Anyone reading a collection of Achmiz's best comments would recognize them as a central example of high-quality rationalist thought, even if they happened to disagree with any given comment. In 2023, Less Wrong moderator Ben Pace wrote, and moderator Raymond Arnold concurred, that "Said is the person independent of MIRI (including Vaniver) and Lightcone who contributes the most counterfactual bits to the sequences and LW still being alive in the world." In 2023, Arnold told me that "there's an important spirit of early LessWrong that [I] keep alive[.]" In fact, we were here first: as far as I can tell, Achmiz and I were part of the community before any of the current moderators. [4]
The relevance of seniority is that, insofar as Habryka agrees that Less Wrong is "less straightforwardly [his] personal fiefdom" because he "inherited what [he] consider[s] a really important cultural institution," one would expect him to agree that we're rationalists just as much as he is—that the cultural institution he's inherited stewardship over was supposed to serve us just as much as people who dislike our writing styles, much as Israel is supposed to be a home for all Jews, even if some Ashkenazi find Sephardic accents annoying. As such, insofar as some people would prefer a Said Achmiz–free Less Wrong experience, one would expect that desire to be met by tools like the existing functionality for authors to ban users from their own posts, which have the great boon of impartiality [5] —not by kicking us off the website we've been using for seventeen years. [6] It is my firm belief that it's possible to share a website with people you disagree with.
The moderation team also used to believe in sharing the website, in the form of an "archipelago" model. It would appear that that era is over. [7] In his self-report, Habryka claims to "care about archipelagos" and that he thinks it's "hard and bad to try to have centralized control about culture", but when I look at his behavior—his comments (four days later, in the same thread!) about the need for "culture to be driven by someone with taste, who trusts their own judgments on matters of culture", his site-wide ban of Achmiz despite the existence of user-level ban functionality as a sufficient and less intrusive remedy, [8] the way he casually threatened banning a user for disagreeing with him—I find the disavowal of centralized control-seeking hard to credit.
One could argue that it's ungrateful nonsense to complain about influence someone obtains via administering a website that they created. The LessWrong 2.0 team led by Habryka built and maintains the current version of the website (albeit seeded with the database of posts and comments from its predecessor), and it was a lot of hard work at great personal expense: of course they're going to run it as they see fit. This "our website, our rules" view may have merit, but it's important to notice that it's a different view than the "inherited [...] a really important cultural institution" view. This section is only arguing that the Achmiz ban is unjustified with respect to the cultural institution view, which may not hold—but if it doesn't, it would be dishonest to motte-and-bailey between the two views, claiming when convenient that LessWrong 2.0 is the fulfillment of Salamon's call for a single locus of discussion for the rationalist community, but insisting when challenged that it's just Habryka's personally curated forum, one of many in the diaspora. [9]
In an insightful post about the perils of delegation, Habryka warns that when you delegate work to someone, that puts them in "a position of leverage over you and the rest of the organization." In retrospect, this would appear to be the devil's bargain the community made when it voted to hand the lesswrong.com domain name to the LessWrong 2.0 reboot. We got a shiny new website as a single locus of discussion—at the cost of the website maintainers usurping cultural authority on the basis of database creds rather than persuasive writing. Was the trade worth it, or would it have better to accept the diaspora as a fait accompli and convert lesswrong.com into a static archive? Given that I'm still here, I can't say with a straight face that it would have been better for LessWrong 2.0 never to have existed. What I can say is that policy debates should not appear one-sided.
III. The Case for the Ban Rests on Misrepresentations
Misrepresentations in Habryka's Analysis of Achmiz's Comments
Habryka presents "a breakdown of the core dynamics that make comment threads with Said rarely worth it" with examples.
As an example of a comment that "read[s] like an implicit claim that you have violated some social norm [...] for which you deserve to be punished (though this will not be said explicitly), or if not that, make you look negligent by not answering an innocuous open-seeming question", Habryka presents a 2018 comment by Achmiz on Benjamin Hoffman's "Zetetic Explanation". In the comment, Achmiz juxtaposes quotes of Hoffman criticizing conventional pedagogy as both lacking "connection to the ordinary means by which [students] navigate their lives" and failing to resemble "the sorts of skills used in time-travel or Robinson Crusoe stories." Achmiz then comments, "Hm."
It's expressed rather elliptically, but what Achmiz is doing with this comment (as he clarified in followup discussion) is implying that Hoffman is in error to present portrayals of skill in fantasy fiction as representative of good holistic pedagogy of real-world skills. Whether you agree or disagree, this is clearly an on-topic, intellectually substantive contribution. [10] The only way I can read it as "an implicit claim that [Hoffman] violated some social norm" "for which he deserves to be punished" is if there's a social norm against making posts that can be criticized, for which the punishment is criticism—but that would be absurd and contrary to the nature of a discussion forum. Arguing is not a punishment!
Habryka cites another response from Achmiz to Hoffman as an instance of Achmiz "dismiss[ing] your response as being totally insufficient, confused, or proving the very point he was trying to make". This is a fair description—but Habryka doesn't address whether the dismissal is justified! Is it not conceivable that someone might write a response that's insufficient, confused, or proving the very point the critic was trying to make?
In the comment in question, Achmiz is disputing Hoffman's claim that the explanation of yeast in the post was adequate for being "competent to interact with the thing, and with its precursors, or at least have an idea of how you'd learn to do so." [11] When Hoffman claimed the post empowers the reader to make hard cider from apples, Achmiz disputed this by pointing to cider-making instructions that claim that naïve methods produce poor results, which Habryka cites as Achmiz "continu[ing] to make insinuations that your failure to respond properly validates whatever judgment he is invoking"—again without addressing the merits of Achmiz's claim.
But the merits matter! If someone believes that Achmiz is applying the wrong standard when he expects that an explanation of yeast adequate to make one competent at interacting with it should allow one to make good bread or cider, that could at least be argued, but Habryka shows no interest in arguing it. He concludes his summary of alleged commenting dynamics by noting that when other commenters ask by what standard Achmiz is invoking a negative judgment, Achmiz explains that he's talking on the object-level and not trying to assign someone low status.
But the comments about the pedagogy of yeast are in fact on the object level about the pedagogy of yeast. One could perhaps infer that when Achmiz doesn't find an interlocutor's replies to be satisfactory, Achmiz is privately making a negative character judgment about the interlocutor as a person, even if Achmiz didn't say that. Such an inference might well be correct, but it's hard to see why it's a moderation issue. On a discussion forum that attempts to grapple with difficult topics at the frontiers of human knowledge, it should not be unusual for commenters to disagree with an author in a multi-comment exchange and not be convinced by the author's replies. (Bayesian reasoners would never agree to disagree, but humans are known to be far from the theoretical ideal.) Part of disagreeing is not finding the other's replies adequate. (If you did find them adequate, that would resolve the disagreement.) Punishing Achmiz and only Achmiz for frequently finding authors' replies insufficient (because people suspect, rightly or wrongly, that Achmiz is privately judging them) would be incoherent, and being coherent about it would amount to plunging full speed into the LinkedIn attractor. Avoiding the possibility that strangers on the internet might be judging you is simply not a realistic or desirable goal for a public discussion forum.
Habryka also claims (in the previous section) that as far as he can tell, Achmiz "refuse[d] to do much cognitive labor in the rest of the thread" on "Zetetic Explanation", but this is at odds with Habryka's own comment in that thread praising a 750-word comment from Achmiz as "a great comment" and saying that Habryka "would maybe like to see this broken out into its own post." Presumably one could not write such a comment while "refus[ing] to do much cognitive labor"! The discrepancy is again glaring.
Misrepresentations in Habryka's Claims of Alleged Author Complaints About Achmiz
Footnote 1 of the ban announcement claims that "many top authors cit[e] [Achmiz] as a top reason for why they do not want to post on the site, or comment here." This is a surprising claim. [12] Is it true? In some sense, it depends on what's meant by "many" and "top", but we can investigate what is known about author sentiments towards Achmiz without immediately pinning down an operationalization.
We do have some direct testimony from some authors. Most prominently, Duncan Sabien (14K+ karma, 16 Curated posts) reported in 2023 that he withheld a post from Less Wrong because of Achmiz, and has written a post denouncing Achmiz's [13] commenting style as amounting to "the willful destruction of gardens of collaborative inquiry." Elizabeth van Nostrand (22.8K karma, 12 Curated posts) said she thought "it would be good if Said left LessWrong." Trinley Goldenberg (5.7K karma, 2 Curated posts) reported "a strong sense of sneer on mine and others['] posts that I find unpleasant" and thanked the moderators for the ban. DirectedEvolution (12.5K karma, 144 total posts, zero Curated) wrote in 2023 that Achmiz is "one of three people who are readily top of mind at having a net negative impact on my LW experience"—but, notably, that "personally I do not favor banning people sitewide for making me feel uncomfortable." Lucas Gloor (4K karma, 5 total posts, zero Curated) wrote in 2023 that Achmiz's questions are "one of the most off-putting things [he] associate[s] with LW"—but, notably, that the user-level ban functionality "seems like a good solution to [him]", and in 2024 added that in the subsequent year he had liked several of Achmiz's comments and didn't remember any that bothered him.
Beyond direct testimony from authors, we have reports from the mod team about private complaints—but, troublingly, as I explain below, previous investigation has shown such reports to be false in their claims about the opinions of at least two authors, which casts some doubt on their overall veracity.
In a June 2025 moderation discussion, Achmiz expressed doubt that authors "capable of making useful contributions to the site" would be discouraged from using the site by the presence of a commenter deemed "'unpleasant'" despite not posting "vulgarity, or personal insults, or anything bad or crazy like that." Achmiz asked for examples of such discouraged authors.
Habryka gave a list of allegedly discouraged authors (bolding added):
I reached out to Scott Alexander via Discord in July 2025 to ask if he had "any specific feelings about Said Achmiz and whether he should be allowed to post on Less Wrong." Alexander issued this statement:
In response to being presented with Alexander's statement contradicting his claim about Alexander's opinion, Habryka replied:
I followed up with Alexander via Discord in May 2026. Alexander said, "I defer to Oli about what we might have talked about in 2019." When asked what defer to meant in that context, Alexander said, "I don't remember and basically trust Oli."
Notably, this is not a corroboration of Habryka's original claim that Achmiz "in particular" discouraged Alexander from using Less Wrong: if Achmiz's comments were so noxious as to drive Alexander off the website, one would have expected Alexander to have at least some memory of it. Indeed, it's conspicuous that Habryka's elaboration does not claim that Alexander volunteered Achmiz's name, only that the "name came up" and that Achmiz "was a commenter we discussed." Given that neither party can recall the details of a conversation that might have happened in 2019 or possibly some other year, which participant is more likely to have first brought up Achmiz: the one who now has "no direct opinion" on Achmiz, or the one who writes 1800-word comments about how he "do[es] just think [it's] false" that Achmiz "does not carry enmity in his heart" "after almost a decade of thinking about it on and off"?
Separately, the inclusion of Jacob Falkovich in the list of authors allegedly discouraged by Achmiz is contradicted by an October 2018 comment by Falkovich (bolding added):
In response to being presented with Falkovich's comment contradicting his claim about Falkovich's opinion, Habryka replied:
In fact, I do not see how the Jacob Falkovich one is "clearly" complicated. Indeed, I dispute Habryka's characterization of what Falkovich "basically says": it is tendentious to paraphrase "negative [...] flipped entirely to become positive" as "made that a lot better." The latter is compatible with Habryka's original claim that Falkovich is discouraged from using the website by Achmiz, but the former directly contradicts it: if something that's bothering you is "made [...] a lot better", the implication is that it's still bothering you a little bit (although not as much as before); if your attitude towards something "become[s] positive", that implies that it's not bothering you.
The significance of Alexander and Falkovich turning out (in their own accounts) not to have been discouraged by Achmiz extends beyond removing them from our list of discouraged authors. It also calls into question the veracity of Habryka's reports of discouraged authors: if two of the names Habryka gave as examples turned out to be non-examples when checked, [15] that, along with other peculiarities in Habryka's claims, [16] suggests that Habryka's reports do not reliably distinguish examples from non-examples and therefore that his claim that there exist unnamed "dozens of others" cannot be taken as reliable. [17]
Overall, while it's clear that some users have some negative sentiment about Achmiz (as we can see from the testimony from Sabien, van Nostrand, Goldenberg, DirectedEvolution, and Gloor quoted above), the support for the more specific and much stronger claim that "many top authors cit[e] [Achmiz] as a top reason for why they do not want to post on the site, or comment here" seems questionable: the only unambiguous example I'm aware of is Sabien.
Habryka claims that "no author complaints were load-bearing for this banning decision" and that he "would make the same decision even if no prominent authors had complained" because he "had much more than enough direct engagement with Said, and seen many more than enough comments of [Achmiz's] on [his] own to understand the consequences of [Achmiz's] commenting style first-hand." That's why the announcement doesn't emphasize author complaints and "instead give[s] detailed models for 10,000+ words[.]"
This position has strange tensions which highlight the dangers of declaring part of a post to be exempt from productive criticism on the grounds of being allegedly "non-load-bearing." Models—especially "detailed models"—are supposed to make predictions! If your model says that the consequences of Achmiz's commenting style are site-threateningly bad, one would naïvely expect that model to predict that prominent authors would complain—such that if authors hadn't complained, or if reports of author complaints turned out to be bogus, that would be probabilistic evidence against the model. If evidence from user complaints is excluded, it's hard to see what consequences of Achmiz's comments Habryka could have observed "first-hand", other than his own personal feelings towards Achmiz.
Misrepresentations in Habryka's Characterization of Achmiz on Obligations to Reply
Habryka claims that "Said himself does not think [ignoring his comments is] a valid option", citing a 2020 comment by Achmiz on obligations to reply to comments asking for definitions or examples, but it's not clear what this adds to the case against users simply banning Achmiz from their own posts: it's weird to cite Achmiz's stance on obligations to reply to support the claim that ignoring him is not an option given that you don't agree with him about obligations to reply.
Moreover, quoting the 2020 comment out of context conveys a mistaken impression of Achmiz's views, which Achmiz clarified in a subsequent comment the same day. [18] In the clarifying comment from the 2020 thread, Achmiz states, "I'm not saying that there's a specific obligation for a post author to post a reply comment, using the Less Wrong forum software, directly to any given comment." Rather, Achmiz explains, the need for a clarifying question to be answered could be met in any number of other ways, such as being addressed in future work by the author, or by someone else other than the author answering the question or linking to an answer elsewhere. The point is not to impose a legalistic burden on the author, but for the question to actually get answered. The "obligation" to reply is meant in the same sense that one has an "obligation" to provide evidence, cite sources, or correct factual errors: it derives its normative force from its truth-tracking function, not mere social convention. [19]
The normativity of truth-tracking is likewise key to understanding the motivation for the 2020 comment's controversial claim that an author who ignores simple requests for clarification "should be interpreted as ignorant." To be sure, there are any number of reasons why an author might not engage with a question other than being unable to answer it (maybe it wasn't worth their time, maybe they missed the notification, maybe they died, &c.), but the relative likelihood of the "can't answer" and "could answer if they wanted to" hypotheses as an explanation for the observed non-response represents predictions about the real world and can't be changed on a whim. A norm that forbid commenters from interpreting the absence of an answer as probabilistic evidence of inability to answer would be requiring them to believe something false. [20]
The enduring misunderstanding of Achmiz's position may owe in part to Achmiz's 2020 comments taking it as an implied premise that the question that demands an answer is a good or legitimate one. In a 2023 moderation discussion, Achmiz clarified that he doesn't believe all comments must be answered regardless of merit:
The reason this 2023 comment is compatible with the controversial 2020 remarks is because Achmiz believes that (what he considers) simple requests for definitions and examples are productive, even if not all comments are productive. According to this view, a request for examples of a phenomenon should not be hard to satisfy if the phenomenon is real; a request for a definition of an niche term should not be hard to satisfy if term means anything. Ideally, the answers should have already been in the post; the fact that commenters have to ask is a manifestation of the post not being complete.
Elsewhere, Achmiz wrote in 2023 that "it's not necessarily (and, indeed, not likely to be) worth your time to engage with all of your critics", and in June 2025, proposed "ignore" functionality (which already exists on the third-party GreaterWrong reader) as an alternative to user bans. [21]
It is troubling that the announcement that Achmiz is being banned in part for his views on discourse norms mischaracterizes those views, despite clarifications in the weeks leading up to the ban: Achmiz explained his position to Habryka in June 2025 (linking to the clarifying followup comment in the 2020 thread), and I highlighted the 2023 "by all means downvote and ignore" comment in a July 2025 message to Less Wrong moderator Ben Pace.
IV. The Ban Was Unnecessary
User-Level Bans Are a Sufficient and Less Intrusive Remedy
Attempting to forestall the objection that people who don't like Achmiz's comments could simply ignore them, Habryka claims that ignoring Achmiz's comments is "not usually a socially viable option." However, since 2018, existing site functionality already permits users with 50 karma to ban commenters from their own Personal Blog posts, and users with 2000 karma to ban users from their own Frontpage posts. If, arguendo, a minority of authors have a legitimate interest in preventing Achmiz from commenting on their own posts, then those authors [22] could already get what they want using existing site functionality, and indeed, were already doing so: we're told that 8 users had banned Achmiz from their posts. Why doesn't that solve the problem without the need for a site-wide ban?
The ban announcement claims that the moderators have told Achmiz "to please let authors moderate as they desire", but this is a curious usage of the word let, relative to how people think about blocking or banning functionality on other websites. Blocks work by the software preventing the blocked user from sending messages. The consent of the blocked user is neither required nor relevant nor expected. [23]
The complaint about not "let[ting] authors moderate" seems to not be about attempts to evade the ban (e.g., by creating sockpuppet accounts), which have not occurred, but simply that Achmiz's comments elsewhere on the website (e.g., on his own shortform) may make users feel less comfortable banning him from their posts. Whether this constitutes not "let[ting] authors moderate" can't be assessed based on authors' feelings alone; the nature of the comments matters.
The offending comments seem to be that Achmiz has argued in moderation threads that the existence of the user-ban functionality enables authors to abuse moderation tools to suppress their critics, noted on his own shortform that he couldn't add further replies to a discussion that he had already participated in before being banned, published on his own shortform a comment that had failed to post because he didn't realize he had been banned, responded on his own shortform to a comment about him, and responded on his own shortform to a comment stating that the author "was honestly curious what you [Achmiz] could mean here".
These are all reasonable comments. It makes sense for users of a website to have opinions about the features that the website supports—and the complaint that the ban feature enables suppression of criticism is just straightforwardly true (even if one might think that it has benefits that outweigh that cost). It makes sense to want to state for the record that one is no longer able to reply to a discussion that one had been participating in, lest someone infer the weakness of one's position from one's non-response. (Indeed, as we've seen, Habryka shares this concern deeply.) Having gone to the trouble of writing a comment and finding that it couldn't be posted, it makes sense to want to post it somewhere, for the effort to have not been a complete waste. If other people are talking about you and your positions in a venue where you can't make corrections (as contrasted to them silently ignoring your existence), it makes sense to want to post the corrections somewhere else.
If Achmiz were using his shortform to harass authors simply for not wanting to talk to him, that would be a matter for moderator concern, but as I've just detailed, the record doesn't support that: every instance I can find of Achmiz supposedly not "let[tting] authors moderate as they desire" is a situation where a reasonable person would have an interest in speaking up. In the case where Achmiz scrutinized another user's ban history, it was contextually relevant to the discussion of how the ban functionality is used in practice; when instructed by a moderator to stop, Achmiz complied.
The user ban functionality works to prevent authors from receiving unwanted comments on their own posts. It's not desirable to prevent banned users from writing on their own shortforms about things that a reasonable person would have an interest in writing about. No further moderator action was warranted.
Effort Ratios Are Logically Irrelevant
In a subsection on "Asymmetric Effort Ratios and Isolated Demands for Rigor", Habryka points out that critics often have an easier job than the authors they critique: asking a question is usually easier than answering it. This is true, because verification is often easier than generation. Anyone trying to make sense of others' writing is going to end up asking questions that are easier to ask than answer. The relevance is not obvious, however. What is the problem here supposed to be, exactly?
Habryka acknowledges that there's no problem "in an environment with lots of mutual trust and trade", but contends that "in an adversarial context it means that it's easily possible to run a DDOS attack on basically any author whose contributions you do not like by just asking lots of questions, insinuating holes or potential missing considerations and demanding a response approximately independently of the quality of their writing."
It would seem that some dubious implicit assumptions are being made about participants' motives. The DDoS attack analogy would carry if we construe intellectual discourse as a zero-sum game in which every unanswered criticism results in the critic winning and the author losing the same number of "points".
But that construal fails on two counts. Firstly, it shouldn't be an adversarial situation. An intellectually honest author (who wants the truth to prevail, rather than wanting their own ideas to prevail regardless of whether they're true) should welcome probing questions. If any of the questions reveals a real flaw in the author's thesis, an honest author would count that as a benefit (because they learned something), not a cost. At the same time, dumb questions that aren't worth answering or reading can simply be downvoted. That is, to the extent that the community can distinguish between good and bad criticisms on the merits, [24] determined critics of a particular author are just providing free verification labor, not conducting a DDoS attack: as it is written of the fifth virtue, "Do not believe you do others a favor if you accept their arguments; the favor is to you." (And to the extent that the community can't distinguish between good and bad criticisms on the merits, we have bigger problems.)
Secondly, even if we suppose that authors are not intellectually honest and just want their own ideas to win, no one should be thinking that an unanswered criticism represents a zero-sum transfer of status from author to critic. Insofar as we model people's discursive behavior as being governed by implied status rewards, clearly the reward for asking a probing question should be less than the reward for answering it (because the answer is more expensive and therefore requires a larger reward to elicit it). Moreover, people are pretty smart about status games and can figure this out. Probing questions rarely get high karma; thorough answers often do. No one thinks that a famous author must be worse than an unknown one because the former attracts many critics and the latter few or none. The market clears.
And given that the market clears, there's no good reason to censor probing questions: whether the questions get answered or not, discerning users will make up their own minds and vote accordingly. The only reason you would want to censor probing questions is to prevent people from noticing flaws in need of probing.
V. The Founding Values Have Not Been Refuted
Lightcone Infrastructure board member and Less Wrong BDFL Vaniver commented on the ban announcement. Vaniver characterizes the dispute over Achmiz as a cultural clash between the values of an older culture of Less Wrong (as represented by Achmiz and the present author) and a newer one (as represented by the mod team). Vaniver concedes that by the standards of the old culture, "pointing out that something is impeding the flow of information is almost enough to end the conversation on its own." He makes an analogy to the development of evidential, causal, and functional decision theories (henceforth EDT, CDT, and FDT, respectively): CDT improves on EDT by ruling out non-causal decision influences that CDT considers illegitimate, while FDT further improves on CDT by ruling in some kinds of non-causal influences as legitimate after all.
Ideological Turing Test–Passing Is Not an Argument
Vaniver then claims that the new culture is the more sophisticated one (analogous to FDT), but the justification he offers is fundamentally lacking if not absent. Vaniver writes:
But that's not what the Ideological Turing Test (henceforth ITT) is for. You can't declare victory in a debate by citing your purportedly superior ITT results: a professor of comparative religion who thinks that all faith traditions are true in their own way would probably do better on an ITT for Islam than most atheists, but this does not imply that atheism is false!
If Vaniver could actually pass old Less Wrong's ITT test and is not just pretending to be wise, he would understand why his purported ability to paraphrase both sides' arguments is not itself an argument, because your strength as a rationalist lies in your ability to argue for a belief being constrained by whether that belief is actually true. As Eliezer Yudkowsky wrote in 2008 (before the term "Ideological Turing Test" was coined): "I cannot argue effectively for that, because I do not believe it. Or if you prefer, I do not believe it, because I cannot argue effectively for it."
Vaniver cites my "Critic Contributions are Logically Irrelevant" as a "crisp" example of the old Less Wrong ethos conflicting with the new. (As many readers inferred at the time, that post was written as a generalized reply to claims that Achmiz's alleged lack of positive contributions justifies censoring his critical comments.) "[P]eople often raise objections about commenters that don't make sense as logical criticisms. But if they aren't intended as logical criticisms, that seems irrelevant to me," Vaniver writes, with a suggestion to reread Yudkowsky's 2007 post "Feeling Rational".
I have to question whether Vaniver read past the title of "Critic Contributions Are Logically Irrelevant": I think I adequately handled the case of objections not intended as logical criticisms in the final section, "A Caveat: Critic Contributions Can Be Relevant If You Don't Care About Maximizing Correctness". The relevance of people having objections about commenters that don't make sense as logical criticisms is that if the objectors have their way, that establishes a precedent of the forum being a place where logical criticism is dismissed in favor of other concerns, which might be desirable for a social or hobby group whose only mission is for its members to have a fun time, but is contrary to the mission of an intellectual forum like Less Wrong purports to be.
Vaniver's suggestion that I reread "Feeling Rational" is a non sequitur. In that post, Yudkowsky argues against the popular misconception that rationality demands suppression of emotions; on the contrary, Yudkowsky argues, it's rational to fully feel the emotion that fits the facts. But I don't argue for the suppression of emotions. Rather, I advocate specifically against the weaponization of emotional harm as a pretext to censor intellectually substantive criticism in a purportedly intellectual venue. If those who find Achmiz's comments hurtful agreed that their pain was not cause for moderator action, there would be no dispute.
Fortitude Is a Rationalist Skill
Vaniver cites Duncan Sabien's concept of "emotional tallness", saying that he used to think of his own psychological fortitude as a rationalist virtue, but now considers it an incidental fact about himself: similarly, IQ is relevant to rationality, but isn't a virtue per se.
But what counts as a virtue is not relevant in the present context. Let us grant without question that no one is morally responsible for their IQ. Regardless, it would clearly be grossly detrimental to the site's mission to ban high-IQ users (or force them write more like low-IQ users) in order to make low-IQ users more comfortable. Or at least, while it's not literally impossible that some sort of special accommodation for low-IQ users might pass a cost–benefit analysis, a detailed case would have to be made for the surprising claim that such an intervention would benefit the site's mission.
The same goes for "emotional tallness." It's a trait relevant to rationality. (Limits on what kind of information you can process without being psychologically impaired imply limits on your ability to make rational decisions in situations where that information is relevant.) A site dedicated to advancing the art of rationality shouldn't go out of its way to silence people with high values of the trait in order to accommodate people with low values of the trait.
It's not clear why Vaniver (or Sabien) uses a height metaphor instead of a word like fortitude or resilience—or if a metaphor for bodily attributes was desired, strength would be the standard one. The strength metaphor would be more fitting insofar as it comports with the reality that it's possible and desirable to increase one's psychological fortitude. It's possible to want to become stronger, and succeed at training it—notably unlike height or IQ, which are largely immutable given current technology. One has to wonder whether the choice of height as the metaphor for mental fortitude is intentional in some sense, meant to protect the self-image of those who don't want to become stronger by making it look like they don't have a choice. [25]
Vaniver concludes by linking to a 2013 comment thread in which he criticizes Yudkowsky's advocacy of deleting comments that negatively reinforce contributor efforts. The present-day Vaniver writes that he "was missing the concept of emotional tallness, then" and that he "can see the younger me levelling a similar criticism at the mod team now."
But the mere observation that present-day Vaniver possesses "the concept of emotional tallness" doesn't explain what the younger Vaniver was allegedly wrong about! [26] In the 2013 thread, the younger Vaniver testifies that "responding appropriately to criticism is a skill that takes development" and that he was "pleased with how far [he had] gotten." Adopting the novel metaphor of "emotional tallness" rather than just saying fortitude doesn't refute the younger Vaniver's claim that it's possible and desirable to learn to respond appropriately to criticism. (Alternatively, if the claim is that it was possible for the younger Vaniver to learn because he was "emotionally tall", but that "emotionally short" people can't learn, it remains to be argued for why Less Wrong should accommodate emotionally short people more than it currently accommodates low-IQ people.)
I, too, have grown and changed a lot over years of life experience, but in the many, many cases where I now disagree with something I used to believe strongly, I expect to be able to explain what arguments and evidence changed my mind in words that my younger self would find at least somewhat compelling. I think that if I couldn't, that would be a troubling warning sign that my current belief was a product of corrupting social pressures rather than wisdom.
In terms of Vaniver's decision theory metaphor, as a convert from CDT to FDT, I expect to be able to give my CDT-adhering ex-comrades a hard time by challenging them to explain why they don't get the big money in Newcomb's problem. If I just said, well, the culture has moved on, people used to believe in CDT, but everyone who's anyone is into FDT now, that would not be compelling.
The fact that Vaniver, while claiming to have transcended the childish concerns of the old culture of Less Wrong, shows so little interest in arguing his views to those who do not already share them is not to his credit, nor to the new culture's.
VI. What I Think Is Actually Going On
In the preceding sections, I have argued that the site-wide ban of Achmiz was unnecessary, contradicts the stated and founding values of Less Wrong, and rests on a slew of factual misrepresentations. The persecution of Achmiz is so bizarre, so indefensible on the intellectual merits, that it seems insufficient to merely argue that it was wrong: the fact that it happened at all cries out for an explanation. Habryka, Vaniver, and the rest of the LessWrong 2.0 team are not stupid people. What could possess them to behave like this—to expect to get away with it, and actually get away with it? Here I am less confident, but a post on the topic would not be complete without at least an attempt to theorize. [27]
Achmiz's Detractors Accuse Him of Passive-Aggression
A recurring theme in the accusations against Achmiz is that Achmiz is allegedly being deceptive, passive-aggressive, or potentially even "gaslighting" [28] by disrespecting other users and then denying that disrespect was intended. Pace, for example, characterizes Achmiz as "holding extreme disdain and disrespect for interlocutors while being committed to never saying anything explicitly or even denying that it is the case."
I can't speak for Said (and he is no longer permitted to speak for himself), but if I may be permitted to speculate, I think what's going on here is that Said is polite. He's a polite person! The reason he's "committed to never saying anything explicitly" is because he believes that personal insults would be contrary to the etiquette of a civilized discussion forum.
What he does do, and what provokes the accusations of dishonesty, is ask pointed questions such that some people will (rightly or wrongly) infer that he's thinking negatively of someone. The complaint seems to be that this is worse, because passive-aggression is purportedly worse than overt aggression. Habryka writes in the comments that he "think[s] [he] could handle disdain fine, if it was carried openly, and could be argued with."
... But It's Hard to Believe That Open Disdain Would Be Better
I really have to question whether openly carried disdain would be better. For myself, I often have negative judgments of people on this forum, but I usually don't articulate them in my comments, partially because that would be rude, but more fundamentally because it would be off-topic. This is a website for discussing topics like probability theory, the philosophy of science, the psychology of decisionmaking, or artificial intelligence. "Which users of this website does Zack M. Davis think are phonies" doesn't make the cut. That's not what people are here to read.
Am I ... somehow wrong about that? Have I been deceiving and gaslighting people this whole time by engaging with the substance of their ideas and trying to keep my thoughts about how stupid and dishonest they are to myself?
Well, unlike my friend Said, I am not a polite person and am not committed to never saying anything explicitly nor denying that it is the case, so if it would help, I'm happy to clarify that I hold extreme disdain and disrespect for the entirety of the Less Wrong moderation team and a decent fraction of the userbase for general reasons that are neatly exemplified by the concrete case of their relentless persecution of Said Achmiz. If it would help, I could elaborate on this for some time using much stronger language.
But I don't think it would help. In June 2025 moderation discussion, Habryka claimed that "[i]t is indeed not a norm on LessWrong to not express negative feelings and judgments" and that "the issue of contention" with Achmiz was "passive-aggression, not straightforward aggression." He then refused to explain what forms of "straightforward aggression" would be permissible, claiming that the question was unmotivated.
But the motivation is obvious: if Achmiz's on-topic, intellectually substantive criticism is not permitted on the grounds that he's allegedly cloaking negative judgments in passive aggression, the implication would appear to be either that such judgments should be made with "straightforward aggression", or that negative judgments are not actually allowed.
It's hard to see what straightforward aggression has to recommend it, however. When an author writes of having "heard of some really awful real-world cases" of some phenomenon and Achmiz replies, "Could you cite some such cases? I think it would be quite instructive to examine some case studies!", it's fair enough to suspect that Achmiz suspects that the author is bluffing and doesn't really have any examples—but what else is he supposed to do but ask? To instead say "I think you're bluffing" (in order to be straightforwardly rather than passively aggressive) just seems obviously worse—not only because it's more unpleasant to receive an accusation than a question, but more importantly because the author might not be bluffing! [29]
In any case, it does not appear to be the case that straightforward aggression is allowed in practice. In July 2025 moderation discussion, I wrote that "I think grown-ups should be able to shrug [...] off" other people disapproving of their behavior "without calling for draconian and deranged censorship policies" and that "[t]he mod team should not be pandering to such pathetic cry-bullying."
Habryka replied:
First of all, Habryka is misusing the term strawmanning, [30] but more importantly, I meant what I said, and I'm happy to explain it in different words if the words I initially used in haste are inadmissible on account of their scornful tone. At issue in the comment in question was Achmiz's stance on obligations to reply. I'm saying that people should be able to tolerate the fact that Achmiz will think less of them if they don't answer questions that Achmiz thinks are productive. If someone's mere personal opinion that questions should be answered is not allowed to be expressed, then that's a harsh and ill-considered policy whose advocates are not behaving admirably because the policy is using people's emotions as a rationale for demanding changes in others' otherwise acceptable behavior.
Now that I've rephrased it to use less inflammatory words, is that still a banned opinion? Is that a thought that LessWrong 2.0 users are not just not allowed to say, but not allowed to think and continue using the website? Because even if I don't say it, I'm probably going to write in such a manner that you'll infer that I'm thinking it, which is all anyone had on Achmiz—and in any case, I'm telling you right now that I'm thinking it.
We would appear to have reached a contradiction: it can't simultaneously be the case that passive aggression is grounds for banning, and that straightforward "aggression and scorn" is grounds for banning, and that it is "not a norm on LessWrong to not express negative feelings and judgments."
The paradox is resolved if we posit that the relevant distinction is not really about aggression, but rather about acceptable targets. Habryka writes that "it is often appropriate to be rude and offensive, and often inappropriate[,]" that "LessWrong has a culture, a lot of which is determined by what things people are rude and offensive towards[,]" and that "[o]ne of [his] jobs as a moderator is to steer where that goes."
On the same day in July 2025, Habryka also wrote that he has "mostly procedural models about how LessWrong should function, including the importance of LW as a free marketplace of ideas[.]" I think it's pretty weird to write both of those things on the same day.
It would appear that the complaint about passive-aggression is a rationalization that can't be taken literally. The only way I can make sense of the observed behavior is to posit that the real offense is possessing [31] negative feelings and judgments towards unacceptable targets. If the person with negative feelings is polite, that gets framed as "passive aggression," but no one honestly thinks that being less polite would be an improvement.
The Real Rule Is to Cover for the Ingroup
So if the complaint about passive-aggression is a rationalization, what are Achmiz's detractors really after? What determines the acceptable targets?
It can't be that they're trying to silence all criticism whatsoever—and not even only because they couldn't get away with that (without third parties noticing). These people aren't insane enough to not understand that being told why you're wrong is useful for becoming less wrong.
Rather, going off of comments that explicitly talk about, for example, "a general trade off between authors' experience and improving correctness", it seems like they're trying to strike a balance between ego gratification and being less wrong. You can get away with making on-topic, intellectually substantive criticisms, but only if you take care to manage ingroup members' emotions and cover for their reputations—as if to put a hand on the other's shoulder and say: hey, buddy, while I'm disagreeing with the literal thing you said, you didn't do anything wrong; I'm still on your side; I'm not judging you; I still care about you.
Said doesn't send those signals. Said doesn't make people feel like he cares about them and isn't committed to believing that they did nothing wrong. He wasn't unique in delivering harsh criticism—other people do that sometimes, too. What distinguished Said was his inflexibility, which is to say incorruptibility. Everyone else knows when to flinch and pull their punches—that is, to lie or prevaricate—when the collective vibes say enough is enough. That's why he had to go. [32]
The people trading off correctness for author experience don't seem to realize why others might consider the trade illegitimate and resist participating. Pace writes disapprovingly of "any commenter being able to demand answers to questions at the risk of the post-author's status[.]"
Well, yes. How did Pace think status worked, exactly? "Status" is just other people's beliefs about you. People are going to update their beliefs about you based on their observations of your behavior. Your status is always at risk. [33]
In a similar vein, Habryka expresses a desire for freedom from judgment while struggling to explain why he wrote such a detailed ban announcement, rather than just admitting [34] that he's personally fed up with Achmiz without trying to justify it:
Habryka hopes that "maybe [he]'ll be a better writer in a few years and can get it across more easily." He won't. Habryka is already a decent enough writer to explain what he means; the reason that any attempt to demonstrate that he and others are not insane is doomed is because the thing that he means is completely unreasonable. People don't owe you a lack of negative judgment. [35] You can't force people to convincingly pretend to like you.
Habryka seems to think he can and should, because it would improve his job satisfaction. In a comment on "grappl[ing] with the concept of 'contempt of court'", Habryka writes that "[t]he cost of someone being a dick to moderators is indeed very high" because "because moderator energy is often the limiting factor for a functional forum", [36] but that "Said is a huge outlier in how much he was contemptuous of any attempts to moderate him, so it might just be less of an issue in the future"—apparently interpreting the phrase "contempt of court" as literally referring to the emotion of contempt. This is a cartoonish misunderstanding of the legal concept. It's a term of art that refers to behavior that interferes with judicial proceedings, like ignoring a subpoena. It's not about protecting judges' morale, such that they don't quit their jobs! [37]
Similarly, Habryka writes that he "would appreciate some courtesy to keep discussion to the principles and decision-level instead of critiques of [his] personal behavior, as indeed much of the cost of moderation is measured in having any moderation-adjacent action be torn apart and be requested to be justified or defended." [38]
I reply: what principles? Every principle I can find in the ban announcement is either immediately contradicted ("Every individual owns their own judgment of you") or grounds out in the subjective "by the lights of the moderators." Asking me to keep discussion to the principles amounts to asking me to pretend that the decision was made on the basis of principles, when I see very little reason to believe that.
Preventing "Relitigation" Is a Pretext for Suppressing Dissent
In the ban announcement, Habryka writes that users who disagree with the moderators can "make a post about how you disagree with some decision we made." He also writes:
This would be great if Habryka were telling the truth that users would in fact be permitted to "make a post about how [they] disagree with some decision [the moderators] made." However, when user Sting made a question post, "Which top authors did Said Achmiz drive away?", Habryka unlisted it (making it only available by direct link) and demoted it to a comment on the ban announcement post, stating, "I do really want to avoid this becoming a recurring topic with lots of disconnected top-level posts doing relitigation", and "I think this is the right choice for stuff like this".
When Sting pointed out that their question met the Personal Blogpost criteria, Habryka replied, "To be clear, we relatively routinely merge and move comment threads (as is common practice on most internet forums)."
The claim that the unlisting was a "relatively routine[ ]" operation is questionable: while I'm aware of some three cases where a derailed comment thread was split onto its own post, an investigation of unlisted posts using the LessWrong 2.0 GraphQL API did not turn up any other cases where a human-authored Personal post from an established user was manually unlisted over the author's objections. [39] The question had achieved 48 karma in 19 votes, suggesting that the voting userbase thought it was an appropriate and high-quality post for the site. [40]
I pointed out that when users make posts on topics that have already been discussed before, the moderators do not demote subsequent posts to comments. [41] What made this topic different?
Habryka replied, in part:
At the time, I pointed out that "have to" isn't warranted: other people have an interest in discussing the ban amongst themselves independently of whether Habryka thinks it's worth his time to reply (to which Habryka agreed).
Separately from the semantics of "having to" respond, however, it's not clear to me that Habryka understands what a stunning admission the phrase "which I do not want to do" is.
The personal convenience of moderators is not a legitimate reason to limit the visibility of content! This should be quite clear if we consider any other topic. It would be obviously corrupt if the moderators wrote a post announcing that the "one-third" answer to the Sleeping Beauty problem is correct and demoted any subsequent posts taking the "one-half" position to comments on the announcement that the answer is one-third.
If the moderators in this hypothetical scenario were to defend their actions on the grounds of their "strongly expect[ing] any discussion situated in that context to go a lot better," one would hope users would ask the obvious question: better for whom? As a matter of human psychology rather than rationality, it makes sense that someone who had sunk 60+ hours arguing that one-third is the correct answer might feel personally annoyed if someone else were allowed to make a top-level post arguing for one-half which emphasized different facets of the context: if readers found the second post persuasive, that might be embarrassing for the author of the first post and even subject them to social incentives to spend even more time replying, which they might not want to do.
But there's little reason to think the first author's personal annoyance has anything to do with what would make the discussion go better in terms of readers having accurate beliefs about the Sleeping Beauty problem. For readers who haven't yet made up their mind on whether the answer is one-third or one-half, the obviously fair policy is that both sides should be allowed to post under the same rules, rather than one side's arguments only being allowed in a context chosen by the other. Indeed, if one side insists on asymmetric rules, that's itself suspicious. [42]
In the comments on Sting's question, I asked if I would be allowed to make a top-level post commenting on the ban decision. In his reply, Habryka wrote that it's "quite important to provide context as to the efforts already exerted on the topic and associated lack of response, as the norms by which many, if not most, readers will interpret the lack of a response is as some kind of admission of wrongness", such that it's "bad form to not provide it in the top-level post[.]"
I have no idea what kind of context Habryka is expecting me to provide. He wrote a post justifying his ban decision; I'm writing a post explaining why both the ban and its justification are bad. In doing so, of course I've linked to the ban announcement and extensively quoted its language: I had to in order to argue against it!
If it helps, I'm happy to clarify that I don't think readers should think I'm right because Habryka might not reply. (He's been very clear that he didn't want to spend more than 10 hours of his time responding to comments on the ban, and I'm happy to provide that context.) I think readers should think I'm right because I've successfully argued in the preceding sections that the ban justification is riddled with logical errors and factual misrepresentations. Whether Habryka chooses to reply or not is his prerogative; in either case, readers will judge on the basis of whatever information is available to them. Similarly, that I chose to write this critique was my prerogative. If the things I have to say are personally inconvenient for someone else, that's not my problem.
If it helps, I'm happy to acknowledge that Habryka already spent 60+ hours on the ban announcement, which evokes sympathy for him not wanting to spend any more time on this affair—but I actually think that context that helps my case, not his. If he spent so much time trying to craft a convincing ban announcement, and ended up with such a poor result (such that I could write so many words credibly critiquing it), [43] I think that speaks to the underlying weakness of his case.
VII. Conclusions and Recommendations for Readers
LessWrong 2.0 Has Forfeited Its Intellectual Legitimacy
I began by saying that this post is not really about Said Achmiz. I meant it. I am not writing this post to the moderation team to beg them to reverse the Achmiz ban. Firstly, that's obviously not going to happen. But secondly, on the object level, the ban itself is not of tremendous importance. If Said had found a new hobby and stopped commenting of his own accord, it would probably take me a while to notice. I wouldn't be upset, and I certainly wouldn't write a 18,000-word post about it.
Rather, I am writing to the world in defense of a principle that I've spent the last decade of my life [44] fighting for: that the truth matters, and specifically, that the truth matters more than people's feelings. If someone makes an intellectually substantive comment on an intellectual discussion forum, it makes sense to object if it's false or irrelevant. It does not make sense to object that the intellectually substantive comment makes you feel bad, because whether the comment makes you feel bad has no bearing on whether the comment is true or relevant. As a grown-up on an intellectual discussion forum, it's not other people's job to manage your feelings.
This website was founded from a sense that more is possible—that a sufficiently detailed understanding of the laws of systematically correct reasoning would enable cognitive feats that would both improve lives today and help humanity confront the machine intelligence transition.
It was understood, at the time, that aspiring to systematically correct reasoning was an unnatural ambition that would require effort and probably psychological discomfort. As it was written in the founding texts:
Correspondingly, the founding texts advised confronting the pain rather than fleeing from it:
That is, in the Sequences era, it was understood that even if you don't have any persistent and demanding critics in real life, you should try to simulate them. It was understood that the process would be unpleasant, but that's it's unpleasant because it involves fighting the human instinct to rationalize, and that the reason to do it anyway is because you care more about being less wrong than about having pleasant emotional experiences.
In the comments on the ban announcement, a commenter pleads for posts from less prominent authors to not be criticized, because "[t]here are selection effects on who stays in the community under these conditions."
Indeed, selection effects are important. In particular, a Less Wrong that bans Said Achmiz is selecting for bullshitters who don't care whether what they say is true or false. Because people who care about whether they're telling the truth want to be held accountable! They want to be criticized! [45] They want people to ask questions about their work! Probably, they want people to ask questions about their work even if they personally won't have time to answer, because they understand the concept of a public comment section in which other people might be able to answer!
Once, people who cared about whether what they say is true and who wanted to become stronger could look to Less Wrong as a beacon transmitting best-in-class information on the topic of that inhuman ambition—a place that really forces you to get your arguments together.
Well, the writing's been on the wall for a while, but the Achmiz ban should make it clear to everyone that that era is over now. It's not just that the culture doesn't train people to think the thought that hurts the most and to imagine what a skeptic would say that would be harder to answer. It's gotten to the point that when a human volunteers to do that work for free, the reigning powers ban them specifically for being too hurtful and hard to answer! Those who would rather the beacon be bright than correct have completed their annexation of Earth's global rationalist community and converted it into a hip Bay Area party-and-AI-discussion scene—a great improvement from their perspective. They're not going to give the beacon back.
What now? As always, that's up to everyone for themselves, but I have a few thoughts.
LessWrong 2.0 Is a Website, Not a Culture or an Authority
In the ban announcement, Habryka invites users dissatisfied with how the site is being run to loudly quit, but adds that "not all things are so bad as to make it the right choice to stop using LessWrong altogether" and that he doesn't believe that agreeing with the LessWrong 2.0 team "on all fronts" should be "a requirement for thinking LessWrong is good for the world."
I think it's wrong to conflate using a website with thinking that the website is good for the world. Internet discussion platforms pose a coordination problem: the main reason to use a website is because everyone else is already using it. I use Twitter (officially called X since 2023) as a discussion platform and as a marketing channel for my writing. That doesn't mean I think Twitter is unambiguously "good for the world." On the contrary, I think Twitter has a lot of negative effects on the world (in the form of sapping attention spans, spreading misinformation, increasing political polarization, &c.). It's just that me unilaterally forgoing Twitter wouldn't make a dent in the harms it does. In the absence of a critical mass coordinating to switch to some superior website, I'm trapped—but because I know that I'm trapped, I can relate to the trap in a sane way.
Namely: Twitter is infrastructure—not a community, and definitely not any kind of intellectual authority. It makes sense to not directly link to Substack in one's Tweets in order to avoid being algorithmically punished, but it should be clear that this is a matter of complying with Elon Musk's whims because he took over the infrastructure (legally, by buying it) and has the hard power to do what he wants, not because there's actually anything wrong with Substack. If Elon Musk were to proclaim that X is a culture, a lot of which is determined by what things people are rude and offensive towards, and that a lot of his job as CEO is to steer where that goes, I think self-respecting Twitter users would laugh in his face. Owning a website does not make you a religious authority.
I think students of human rationality should relate to LessWrong 2.0 the same way as they relate to Twitter—as infrastructure—and relate to the moderation team the same way they relate to Elon Musk. [46] I think LessWrong 2.0's suppression of intellectually substantive discussion in order to protect contributor feelings is doing harm by deceiving the world about what rationality even is, but it still makes sense to use the website as a discussion platform and as a marketing channel for our writing, because we're trapped: coordinating to move to a new website is hard enough that—for now, if things don't get worse—it makes sense to stay put and keep doing our thing under Musk's terms of service. (Although it's probably advisable to keep your own blog that you control and merely linkpost to this website as a marketing channel, rather than trusting LessWrong 2.0 to host the canonical copy.) [47] In that sense, I agree with Habryka when he writes that "for the rest of us" "not that much has to change."
However, unlike the case of Twitter, I worry that too many students of human rationality aren't modeling themselves as trapped and therefore aren't relating to the trap in a sane way. I worry that a lot of people are clinging to an implicit assumption that there is a healthy and trustworthy "rationalist community", such that trying to gain the approval of "the community" and its leaders amounts to the same thing as trying to be rational.
That was never going to be real. Those who seek the Way have to invent it for themselves.
Speak the Truth, Even If Your Voice Trembles
To the people who care about whether what they say is true, I advise courage. The people playing ego-flattery and reputation-protection games will hate and fear you, as they hated and feared Said Achmiz. When one of their own declares that beliefs about dragons don't require evidence, they might let it slide when you contradict them and say that all beliefs require evidence. (Mere disagreement doesn't threaten their self-image, as long as they can frame it as coming from a place of mutual respect.) But they won't take it lying down if you try to further point out that the most plausible reason why someone would claim that beliefs about dragons don't require evidence is in order to protect false beliefs about dragons. (That threatens their marketing story about being a community of collaborative truthseekers.) They'll say you're sneering, heaping scorn and social punishment on the virtues and values that are the lifeblood of the community by their lights.
Don't be intimidated. You don't have to comply with their demands to soften, obfuscate, and lie. Just keep speaking clearly on whatever channels are available to you.
To be clear, it's not that these people can't hurt you in any way. They can slander you and silence you on the platforms that they own and ban you from their events.
But in the long run, their power is self-limiting. In a conflict between people who care about whether what they say is true, and people who say whatever they need to say to achieve their immediate social goals, the latter will tend to achieve their immediate social goals—but only by flooding the zone with nonsense that doesn't stand up to basic logical scrutiny and trying to prevent anyone from applying basic logical scrutiny (perhaps under the guise of preventing "relitigation" or "fights of attrition"). They can martyr the occasional principled person when they have an overwhelming local resource advantage, but their reputation protection scheme will never be safe unless they can successfully purge everyone capable of applying basic logical scrutiny—and they need our eyeballs for their platforms and our money for their hotels. Though trajectories are harder to predict than endpoints, it's reasonable to expect that the charade can't last forever.
A similar process played out again during the construction of the present post. The original title of the post was to be "Said Achmiz in His Felon Stripes Stands Far Above You Now", from an adaptation of Voltairine de Cleyre's impassioned 1890 defense of the jailed proto-feminist publisher Moses Harman that I wanted to use in the post:
At first it was going to be an epigraph—to the last section, if I couldn't get away with making it an epigraph to the whole post without getting dismissed by an unsympathetic audience. Said thought it was "too overwrought" and that it remained to be seen if a later draft might "earn it." I couldn't bear to cut it and ended up using it as a peroration in a later draft. Said didn't buy it, saying that his "previous comments about the tone stand" and that it "muddle[d] the concluding section." He was right. ↩︎
Although curiously, Habryka's detailed description of the sneer attractor doesn't particularly seem to resemble Achmiz's comments in any aspect other than them both involving harsh criticism. Habryka writes that "[t]he key component of making good sneer club criticism is to never actually say out loud what your problem is." But Achmiz is quite willing to say out loud what his problem is, and is clearly committed to answering clarifying questions at length if his initial complaint was unclear. ↩︎
The Zionism analogy for re-centralizing a dispersed community writes itself given the term "Less Wrong diaspora"; no particular geopolitical implications are intended. ↩︎
My first comment—while the original Sequences were being written on Overcoming Bias, more than a year before Less Wrong was founded—dates to December 2007; I attended the first Overcoming Bias meetup in February 2008. Achmiz's first comment dates to May 2010.
In contrast, Pace "read The Sequences [...] in 2011", Arnold's account dates to September 2010, Ruby Bloom "found LessWrong in 2012", Ronny Fernandez's account dates to 2011, Vaniver's account dates to October 2010 after he "found [Less Wrong] through a link to HPMOR on the xkcd forums", and Habryka's account dates to 2015 with Vaniver confirming that Achmiz has "been here longer" than Habryka. ↩︎
User-level bans or ignore functionality—and indeed, karma voting itself—are impartial in the sense that everyone has the same tools: if Alice doesn't like Bob, she can downvote and ignore his comments or ban him from commenting on her own posts, and vice versa. The impartial tools help people share the website without the moderation team needing to take sides. ↩︎
It's notable that Yudkowsky's articulation of "Coherent Extrapolated Volition" includes the clause "where our wishes cohere rather than interfere", such that "[a] minor, muddled preference of 60% of humanity might be countered by a strong, unmuddled preference of 10% of humanity." Analogously, a majority of rationalists mildly disliking having to share the website with us shouldn't outweigh our strong desire to share it.
The coherent-wishes clause also illustrates why objecting to banning Achmiz is not tantamount to demanding no moderation whatsoever. No established users are writing 18,000-word manifestos complaining about the rejected posts slushpile being censored, because our wishes cohere on censoring the slushpile in order to maintain the signal-to-noise ratio. ↩︎
In the comments on the ban announcement, Vaniver writes that the correct response to differences between cultures of aspiring rationalists is:
Does Vaniver assume that people on one side of the conflict aren't even interested in reading anything by people on the other side? (One would have imagined that the eleventh virtue of scholarship would recommend being well-read.) Why not just share the website, since we're not going to run out of paper? ↩︎
See the subsequent section, "User-Level Bans Are a Sufficient and Less Intrusive Remedy". ↩︎
An analogy could be made to Habryka's 2024 complaints about the effects of Open Philanthropy (now Coefficient Giving) on AI safety work in which he wrote that "[i]n as much as [Open Philanthropy] is the most powerful actor in the space, the original geeks are being thoroughly ousted." Coefficient Giving might well reply, "our money, our rules"—that it's ungrateful nonsense to complain about how Dustin Moskovitz chooses to spend his own money—but this would seem to be missing the point. ↩︎
The rhetorical device of juxtaposing seemingly contradictory passages from an author with minimal additional commentary as a prompt for the author to clarify, is not a unique innovation of Achmiz's. In a June 2023 email discussion, I received an email from Hoffman himself that quoted two parts of my previous message, with the comment, "?????". The meaning was quite clear. I did not take offense or interpret Hoffman as claiming that I deserved to be punished. I just said, "Thanks for pointing out the tension between those two passages," and explained what I meant.
And what Achmiz dishes out, he can take. In a May 2026 discussion of the æsthetic merits of the musical Hamilton, I perceived an apparent contradiction between two things that Achmiz had said and quoted them with the comment, "Hmm ..." He didn't interpret it as claiming that he deserved to be punished, either. He just said, "Yes. Exactly," and explained what he meant. ↩︎
As an accomplished baker, Achmiz is in a credible position to tell. ↩︎
"Top authors" receive lots of comments by virtue of being "top", and not all comments are good; it would be strange if one commenter merely asking pointed questions—not being abusive or overtly insulting—was somehow so offensive that "many top authors" found him in particular a "top reason" to not use the site at all, particularly given the existence of user-level ban functionality. ↩︎
Not by name, but the discussion in the comments section makes the inspiration quite clear. ↩︎
Presumably Habryka meant the history of the LessWrong 2.0 reboot. Less Wrong launched in February 2009. ↩︎
In May 2026, I also emailed Eliezer Yudkowsky to attempt to corroborate Habryka's claim about Yudkowsky's opinion of Achmiz, but Yudkowsky could not be reached for comment. I'm only aware of one direct interaction between Yudkowsky and Achmiz on Less Wrong, a 2022 thread about the faithfulness of Yudkowsky and Lintamande's collaborative fiction Planecrash to Pathfinder rules as written. ↩︎
Habryka also writes, concerning Benjamin Hoffman, that the conflict between Achmiz and Hoffman on "Zetetic Explanations" and another conflict between Hoffman and Duncan Sabien "seemed like roughly the thing that caused [Hoffman] to leave"—but those incidents were in May 2018 and August 2018, respectively, whereas Hoffman went on to make more than 20 posts (including two Curated posts) between August 2018 and June 2020 (before a gap until April 2022). The lag from putative stimulus to response of almost two years calls into question the reliability of Habryka's ability to perceive what is "roughly the thing that caused" what. ↩︎
The unreliability is implied by probability theory. The evidence of a report of a discouraged author is determined by the likelihood ratio, . If not-discouraged authors are incorrectly reported as discouraged, it follows that any given report is less meaningful. The conclusion is agnostic as to the cause of the false reports: we don't actually need to know whether the false statements were due to dishonesty, ignorance, poor memory, weak reading comprehension, &c. ↩︎
Habryka links to the clarifying comment in footnote 11 as "further details on what Said means by 'responding'", but I don't think this suffices to clarify Achmiz's true views to readers of the ban announcement. ↩︎
In personal correspondence with the present author, Achmiz re-stated his position again:
↩︎More precisely, to assign probability distributions with a worse proper score. ↩︎
The distinction being that auto-collapsing comments from an ignored user doesn't prevent other users reading the comments. Discord's block functionality in shared channels works similarly. ↩︎
Modulo the karma limits, but the behavior of the mod team since the feature was introduced seems to imply much more concern about the harm done by unwanted comments than by frivolous bans: I'm not aware of anyone punished or warned for banning too many users. This suggests that the karma limits could be dropped or revised without issue. ↩︎
The point of a unilateral block feature is precisely that you shouldn't need someone's consent to get them to stop talking to you. ↩︎
And to the extent that readers are taking into account the selection effect of who the determined critics target. ↩︎
Thanks to Said Achmiz for this observation. ↩︎
Vaniver also writes that his younger self was "missing the point about the conversation quality being worse because of indirect effects." This is too vague for me to address without further explanation; I don't know what purported indirect effects Vaniver is referring to. ↩︎
Throughout this part, claims about motives should be understood in a functionalist sense: an agent that systematically chooses actions that result in X can be said to "want X" without necessarily consciously verbalizing, "I want X." ↩︎
In the comments on the ban announcement, Habryka writes that he "feel[s] some temptation to try to restore some of [his] sanity by providing compelling demonstrations of what has felt like gaslighting to me", although that that wasn't the primary motivation for the ban announcement post. ↩︎
A question can simply be answered, but accusations need to be defended against—and if successfully defended against, the accusation is recognized as unjust. This makes accusations a uniquely terrible way to elicit examples in situations where examples would be helpful: the elicitation succeeds if and only if it shouldn't have happened! ↩︎
Strawmanning is mischaracterizing a speaker as having said something other than what they did. Hostile descriptions of someone's behavior (that are understood by readers to not represent that person's opinion of their own behavior) are not strawmanning. In context, I'm obviously not saying people are advocating policies that they think are "draconian" and "deranged", or even necessarily constitute censorship as they would use that word: the negative connotations of censorship make people reluctant to use it to describe instances of information suppression that they approve of. Habryka seems to understand this point in other contexts, as when he writes that "trying to force every description of every organization to pass their [Ideological Turing Test] is bad." ↩︎
In light of the ban announcement's declaration that "You are at least somewhat responsible for the subtext other people read into your comments, you can't disclaim all responsibility for that", I think possessing is the correct verb here, not expressing. Not saying anything negative won't save you if people can infer that you're thinking it. ↩︎
In 2018, Benjamin Hoffman wrote that Achmiz was "one of the few people to consistently call out" "a tendency to conflate the stated ambitions and actual behavior of ingroups like Rationalists and EAs, when we wouldn't extend this courtesy to the outgroup, in a way that subtly shades corrective objections as failures to get with the program."
Achmiz's first moderator warning on LessWrong 2.0 was for a comment answering the original post's question about the lack of popularity of the Center for Applied Rationality's "Double Crux" technique, in which Achmiz stated that the technique's origin at CfAR was "an anti-endorsement."
The warning stated that "A simple 'I think that...' or '...for me' would have done a great deal to resolve" the problem of Achmiz's answer "generat[ing] a lot of unnecessary friction"—but as Achmiz points out, the comment in question already begins with "My take". Was that not enough of a personal-opinion-only indicator, such that Achmiz should have somehow known to also append "for me" to the "anti-endorsement" claim? In the comments on the ban announcement, Habryka cites the fact that the original post had a "Thinking out loud" epistemic status line as evidence that Achmiz would ignore author requests to not criticize an undeveloped idea. But the comment in question wasn't critical of the post, but rather of Double Crux and CfAR—unless one construes criticism of Double Crux to imply criticism of the post insofar as the author writes that Double Crux "failed to take root as well as it should"?
Looking at the pattern of moderator behavior, it's hard to believe that the problem is Achmiz's conduct, rather than him refusing to conceal his negative opinion of CfAR. ↩︎
It has to be, in order to mean anything: in order to gain status when you do good things, it needs to be possible to lose status when you do bad things. ↩︎
Habryka writes that he "do[es] think a huge fraction of this announcement should be seen as 'look, it's been too long, I've spent a lot of energy on this, I am not dealing with this anymore, let's part ways'." ↩︎
Much as Habryka doesn't owe people free meetup hosting space. ↩︎
The moderation page lists comments that have been marked as being made in the role of a moderator. There were two such comments in April 2026 and five in March 2026, which casts some doubt on moderator energy being the limiting factor for LessWrong. ↩︎
The fact that "contempt" in the judicial sense is not about judges' or perpetrators' feelings is immediately clear from a cursory examination of famous "contempt" cases. In United States v. Shipp (1906), a local sheriff was found guilty of contempt of court for allowing a lynch mob to murder a death row prisoner despite a court order granting an appeal. In Miller v. Davis (2015), a county clerk was found guilty of contempt for refusing to issue same-sex marriage licenses. In Illinois v. Allen (1970), a defendant was found guilty of contempt for persistently interrupting proceedings, threatening the judge, and declaring, "There is going to be no proceeding. I'm going to start talking and I'm going to keep on talking all through the trial"—that is, the point was that he was actively disrupting the trial, not merely "being a dick" in the judge's subjective opinion.
The analogy to forum moderation is straightforward: a user who actively defies clear moderator orders would be guilty of contempt in the judicial sense, but Achmiz clearly hasn't done any such thing: indeed, Habryka reports that he "decided to not actually check the 'ban' flag on Said's account, on account of trusting him to not post and vote under his account"! (If technical limitations prevented a software-imposed ban, posting and voting in defiance of moderator orders would be contempt, and Habryka trusts Achmiz not to do that.) ↩︎
A footnote clarifies that "of course in as much as something seems egregious, you and others should feel free to call it out"—but Habryka doesn't seem to consider that someone might find the request itself to be egregious. The only reason to silence discussion of conduct but include an escape hatch for "egregious" bad behavior would be to provide a license to indulge in less-than-egregious bad behavior. Indeed, one must wonder how Habryka himself would respond to Sam Bankman-Fried or Dario Amodei requesting the same "courtesy" from the people they have power over. He can't say, "That's different because I'm the good guy." That's their line!
The asymmetry in who is entitled to such "courtesy" is also striking: Said was clearly not exempt from critiques of his personal behavior! As stated, the argument seems to be that people with power should be subjected to less scrutiny than everyone else, because their jobs are just so hard. ↩︎
A May 2026 query of
query { posts(selector: { unlisted: {} }, limit: 200) { results { _id, title, pageUrl, postedAt } } }found 80 unlisted posts, whose reasons for unlisting were analyzed with GPT-5.5 Thinking. The results included: 11 posts belatedly identified as in violation of the LLM writing policy that were unlisted but not unpublished in order to keep comments accessible (1 2 3 4 5 6 7 8 9 10 11), a clue for the 2024 Less Online puzzle hunt, test posts titled as such (e.g., 1 2 3), posts serving as meta information pages about the site (e.g., About, Contact Us, Privacy Policy, "Quick Guide to Tagging"), appendices of mathematical proofs (1 2 3 4 5 6 7 8 9 10 11 12 13), &c., but nothing that seemed comparable to the unlisting of Sting's question. Thanks to Robert Mushkatblat for GraphQL assistance. ↩︎Moderators do routinely prevent inappropriate or low-quality posts from being published on the site, as can be seen in the rejected posts slushpile on the moderation page, but Sting's question is clearly not in the slushpile reference class. ↩︎
For example, no one seemed to have a problem with "Forecasting is Way Overrated, and We Should Stop Funding It" prompting the reply posts "Forecasting is Not Overrated and It's Probably Funded Appropriately" and "Comment on 'Forecasting is Way Overrated, and We Should Stop Funding It'". ↩︎
If one-half is in fact wrong, wouldn't you expect readers to find your post more convincing on the merits? Crushing them with your mod powers only invites doubt that you weren't convincing on the merits. ↩︎
I think some readers may be wary that a long post such as this one represents an effort to "filibuster": that the mod team did something I disliked, so I'm trying to delegitimize them by throwing up an impressive-looking flood of verbiage. I'm denying this on the grounds that it wouldn't work: that would be bad writing, and the comment section would notice. The reason this post is long is because the reasoning offered for the ban is so terrible in so many different ways that I actually have this much to say about it. ↩︎
Given the existential risk situation, perhaps in more ways than one. ↩︎
Because if they can't answer the criticisms, they learn something, and if they can, they gain more justified confidence in their ideas: an idea is more likely to be right if someone tried to prove it wrong and failed—but for that to happen, critics have to be allowed to try. ↩︎
Who, notably, also talks a good game about "truthseeking." ↩︎
Prior to the ban, I used to write a lot of "Less Wrong exclusives", but now I only intend to do that when the content is specifically about "the community" rather than general rationality or AI topics. I regret that "Terrified Comments on Corrigibility in Claude's Constitution" and "Dispatch from Anthropic v. Department of War Preliminary Injunction Motion Hearing" initially going up as "exclusives" due to technical issues with an old WordPress setup. ↩︎