Me too. I gave myself a 9 last year, and don't feel guilty about that. I think that's the rating I would have given the review if it had been written by someone else.
Am I mistaken, or were you not upset about the Alpha School reviewer getting people to come here and vote particularly and only for his work in the last review? What is the difference in principle there between "I'll vote for myself in this contest because I think it's so good" and "I'll get all my friends and family to vote for myself in this contest because I think it's so good"?
Regardless of one's views on self-voting and on the Alpha School review, I think there's an obvious difference between a single, established member of a community casting one vote for their own review on the grounds that they thought it was that good, and someone attempting to get people from outside the community to support a review they might not even have read for purposes external to the competition itself.
I'm not personally saying that's what happened, but it's obviously Eremolalos's view and I see no hypocrisy here.
I'm not saying it's hypocrisy, but why imagine that the Alpha School author was doing something bad, if he felt his review truly was excellent and deserving of a high finish, in getting votes?
I do think there's a thin line between voting for oneself and gathering up others to vote for you, and it's not a line we should be crossing.
Just to be clear I didn't vote *only* for my review. I read and rated a dozen or so by other people. There had been some discussion of whether we were allowed to rate our own review, and the upshot was "there's no rule, do as you like." I thought, and still think, I was able to fairly apply to my own essay the same rubric I had to others. And by the way, 9 wasn't my highest rating. I gave a 10 to somebody else, and also a 9 to one other person. (My favorite essays were the one called something like "The World -- a Review" and
something like "The Soul of a Right-Wing Man." I thought both reached peaks of insight and wit higher than mine. I was dazzled by them. I quoted favorite lines from both of them here during the voting period, to give other people the pleasure or reading them.)
Having friends and family vote for me seems entirely different to me. First of all, they would all be doing just the one rating, of my essay, presumably giving it a 9 or a 10. They quite likely wouldn't even read it, and even if they did they wouldn't be reading other reviews and giving them a fair shot at getting high ratings too. And they wouldn't be giving mine a high rating because they thought it was good, they'd be giving it because I'd asked them to.
So I can sort of see objecting to someone giving their own piece a high rating -- though my rating was an honest one -- but it doesn't seem to me to be at all akin to sending friends and family here to vote for my essay. As a matter of fact, I sent copies of my review to a couple of friends, who seemed to enjoy it. But it never even occurred to me to ask them to come here and give it a vote. And if I had, they'd have been offended by the idea. Like me, they're grown-ups with well-developed consciences.
You are only one person. I don't like the idea of voting for your own review, but if you do so, you take the total vote count from n to n+1, which is very unlikely to affect the final result.
This is very different from asking your 50 friends/acquaintances/relations/employees/whatever to vote for your review.
So basically, I think one is less bad than the other on consequentialist grounds.
This left me smiling a lot longer than funny comments usually do. I admire the restraint in keeping to a concise eight words and letting the absurd and not-so-absurd implications sink in.
I don't do it. I'm not capable of judging the quality of my own work: some stuff I thought were bad other people really liked, and some things I thought were brilliant landed with a thud.
Without a hard rule against it (and maybe even with one), some authors surely will vote for their own reviews. So it seems like encouraging everyone to vote for their own is a better way to level the playing field than discouraging it. NB I didn't write one.
It feels like "Yes, self vote" is a more stable rule than "No, don't self vote". If nobody votes on their own then defecting gives you a (small) advantage. If everyone votes on their own then there's no advantage to be had.
"The Last Temptation of Christ" only has five paragraphs and appears to just be the introduction - I'm not sure if the author only submitted a partial review or if the issue happened somewhere else in translation to website.
There's another one which is very short, it only seemed to be getting going and then stopped dead. So I marked that one down.
You don't need to produce Gibbons' Decline and Fall in review length, but a bit more than "so here's this book, this is a very brief summary of the plot, yeah I liked it" would be nice.
The review of “The Gods Themselves” seems to have a large chunk that looks like a footnote, perhaps because Rob’s software was derailed by the author’s inclusion of a numbered list at that point in the text.
My review has some minor formatting issues around quotes. Nothing serious but it is noticeable. If you're accepting that kind of feedback let me know how to send it to you (I can't be more specific here as that will link this account to my review!)
Some goes for anyone else who sees any formatting issues with their essays.
I tried to make the import pipeline robust to all the different ways people were formatting stuff, but there was a lot of diversity in markup so it probably has some cases where it's not doing quite the right thing.
Tale of Genji has a misplaced list. The list starting with "He hears a rumour that a woman is particularly beautiful" is supposed to come after "Here is a non-exhaustive list of reasons why the men in Genji fall in love"; instead, it's nested inside another list beginning with "An empress seeks to solidify her son’s power by getting him married well, and works hard behind the scenes to secure a specific young lady as her son’s wife".
This isn't necessarily an error. There are usually a few unfinished submissions. I'm guessing this happens because some people submit the link to their doc early in the submission window intending to finish but run out of time before the deadline.
I'm only one reader but I do start to penalize ones over ~10k words if it seems like they could have made their point without such an expansive length. Some of them seem to use prolixity as a sort of proof of work, but editing down your draft is the most considerate work of all! As the old chestnut goes, "I would have written a shorter letter, but I did not have the time."
"The nearly 10,000-word length will feel excessive to some readers"
That was funny, though one of the reviews I read mentioned something about a book being 380 pages long so probably those reading the review wouldn't bother with the source since it was *so long* and I had to shake my head over the lack of stamina of kids these days. Some of us cut our teeth on "The average Stephen King book is roughly 450 to 500 pages long, containing between 150,000 and 175,000 words" or even longer books, my friend!
If it's lively, interesting, and informative, then no length is too long. If it's bad, repetitive, meandering, or dull, then no length is too short.
The longer it is the more interesting and / or well written is has to be, is how I feel it works. There's just so much time worth spending on "that's mildly interesting".
At least books have the upside of a plot and characters you can get invested in. "I really want to know how it ends!" or "I bet the finale is going to be great" is not a thought I had about, for example, the land tax review that won a couple of years ago.
I liked the review very much, but the contents of the book made me shudder. And then laugh, because yes, the reviewer does gently point out the contradiction between "anyone can do this! it's easy!" and "we have full time childcare because me and my wife work, so we are in the same position as 19th century upper-middle-class who leave all the child raising to the nanny".
That did leave me with the feeling that maybe the childminders are doing the "carrots first, *then* ice cream" discipling and instilling of habits that the author of the book is not doing, and wondering if the amount of interaction with the kids is limited to "two hours every weeknight between us coming home from work and them going to bed. plus weekends" (and I don't even know if the childminders work weekends) and irresistibly made me think of Hilaire Belloc's spoof on "The Nordic Man":
"The Nordic Man has a nurse to look after him while he is a baby, and she has another domestic at her service. He has a night and a day nursery, and he is full of amusing little tricks which endear him to his parents as he grows through babyhood to childhood."
That makes them the fun parents, and it's easy to raise kids when you're being the fun parent(s) and somebody else is doing the no-fun part of discipline and rules. "I let my five year old eat only ice cream if that's what he wants!" is much less harmful if, in the background, there is "and while I'm not there for the other eight hours of the day, Consuelo (our treasure of an au pair) gets to have the fights over eating vegetables and washing his face".
It took me a while to understand that ice cream here is not supposed to be the stick to go with carrots but that, ironically, the stick in this scenario is carrots.
---
As for the topic itself: The review, the book under review and Scott writing on taste seem to believe that children are born with a (not surprisingly simplistic) taste and that we should treat that as a deeper truth of the universe.
I think that is absurd.
If you observe that a child, when given the choice, prefres oreos over strawberries you have learned something about oreos, not about the child. And crucially, if you never give it oreos, you will find it enjoys strawberries as much as the next child enjoys oreos.
To look at that state of affairs and conclude that the child is denied its natural preference for oreos (review) or that a taste for strawberries must be fake (Scott, if only by extension) ... I don't buy it.
The review is almost intriguing enough to make me want to read the book, but I know reading the book itself will want to make me punch the author in the snoot:
"Aaron Stupple's five-year-old son has been eating ice cream "almost exclusively" for the past few months. Prior to that, his staple was Oreo cookies. As a caring and conscientious father, Stupple ensures there are copious amounts of junk food available to his children at all times, strategically stocked in easy-to-reach cupboards.
The first few sections of Stupple's 2025 book, The Sovereign Child, contain such a rich bounty of potential ragebait that I agonised over selecting the juiciest bits for your titillation. Here's his stance on childhood nutrition:
Should parents really just let their kids eat whatever they want? The short answer is yes."
I do wonder about the "almost exclusively", since surely the child has not been living on ice cream? That is vague enough that it could be "he eats ordinary food but when he wants ice cream he gets ice cream, at any time, as much as he likes" up to "yes, three out of four meals are ice cream only".
"A harder question is whether TCS is a luxury available only to those with lots of spare time and money. Here I present to you two quotes from The Sovereign Child, which happened to end up next to each other in my notes, and which made me laugh out loud when I re-read them in sequence:
The self-sovereignty we might unlock for children can take root in any home, anywhere, by anyone of any financial status.
We have full-time caregivers for our kids during the day while my wife and I work."
And that does make me strongly suspect the childminders are doing the bulk of the "make sure the kids are toilet-trained and not coming down with beri-beri" work while Dad and Mom preen themselves on how non-coercive their parenting style is (for the two hours a day they actually interact with their kids).
At least the reviewer has a sense of humour and proportion about all this.
I hadn't planned on reading the review of The Sovereign Child, but Deiseach's response was intriguing enough that I went and read the whole thing. Stupple's view on child rearing doesn't convince me (Deiseach's singling out the bit about the author and his wife relying on full-time child care did that), but I do appreciate his strenuous efforts to appear reasonable and sensible, even in those cases where he has to really strain every nerve.
The quotation I enjoyed the most was when Stupple comes to the conclusion that "there are some cases where hard constraints [on kids] might be justified". And exactly what cases might those be? Heroin? Sex with adults? Carrying guns to school? No. "In the extreme", writes Stupple, "we might need to step in to stop them drinking alcohol, or watching hardcore pornography, or reading Thomas Piketty's Capital in the Twenty-First Century." No knowing the author's ideological background, I can't decide whether the inclusion of Picketty was a misfiring joke or an indication of serious belief. Either way, Piketty would no doubt be amused by it.
I have no intention of reading the book, but I do wonder if Stupple isn't being deliberately provocative; I find it very hard to imagine he lets his five year old eat *only* ice cream or *only* Oreos for weeks at a time.
His potted bio from some institute he is involved with:
"Aaron Stupple is a practicing physician and father of five in Western Massachusetts. He has been promoting critical rationalism and the work of Karl Popper and David Deutsch since 2019 in the form of online community building, a web magazine, and Rat Fest, the annual in-person conference in Philadelphia. Aaron wrote a book on Taking Children Seriously, which is the application of critical rationalism to parenting."
His co-author is also involved with said institute, and appears to be a Libertarian, so I imagine Stupple's politics incline in that direction too:
"Logan Chipkin is a writer and editor in Philadelphia. He currently serves as Managing Editor of a nonprofit that works to spread liberty in the West. He has written articles about fundamental physics, economics, history, and Popperian philosophy for outlets such as Gizmodo, Physics World, Quillette, History Magazine, The Libertarian Institute, The Pennsylvania Gazette, and Bitcoin Magazine, and he has written a fantasy novel called Windfall. He has also edited a book written by a prominent physicist and has collaborated with other physicists to communicate their ideas to a general audience. Finally, he hosts and produces Rat Fest, an annual in-person philosophy conference in Philadelphia."
Interesting professional profile for Chipkin. It is genuinely difficult for an outsider to accurately gauge the value of publication in such periodicals as Gizmodo, The Libertarian Gazette and Bitcoin Magazine, none of which I have ever heard of. They rather sound as though they have an axe to grind, but that may very well be nothing more than an indication of my own relative ignorance.
Well, after 7 reviews I was starting to worry that I was being too stingy with my scores, but I now see none of the ones I read made it into the predicted top 10...so I feel somewhat vindicated. The LLMs, as always, work hard to please us.
"ACX readers will likely find this a solid above-median review ..." and similar statements appear very often even though the LLM in fact ranked the books below median.
although i do kind of agree with the sentiment of not taking a look at the AIs opinions until you've done your own 10 votes or so first, to not bias your choice of what to review.
(also just did a first pass with Sonnet, then using Opus 4.8 on the top 75 for, going to have 4.8 go through the rest after my limit resets)
Pretty interesting, thanks. I agree it would be a shame to check before doing a decent bit of reading first.
There's one I liked that it doesn't rate at all, and a couple of the ones it rates very highly I thought were a notch or two lower, but as a general scan it works well. I think its analyses of the strengths and weaknesses are pretty spot on, the issue will just be how accurate it is in the weighting with regard to overall score.
Thank you for making the fancy book review website! A big improvement, quality-wise, from the giant Google Doc of old!
I'll signal-boost other readers' comments: would you be able to add a way for readers to post comments on the reviews, including ones that won't be posted on ACX? People put a lot of effort into these reviews, and I'm sure they would appreciate some constructive feedback.
Sorry for the silence. I've been silent because I'm still thinking about what's best here.
As you say, the pro is that a lot of these reviews are really good and it would be nice to allow people to talk about them and be able to connect with authors.
Challenges are:
* Unlike when reviews are posted to the blog, reader attention is getting split across 162 reviews, so discussion probably wouldn't be great on individual reviews
* Moderating 162 separate conversations will be a pain
* Building a commenting space that actually works really well is non-trivial (I've worked on comments at multiple employers)
* It could be hard to maintain anonymity
I'm also somewhat interested in the "form to send email to the review author" idea too. Though we'd need to think about how that fits with anonymity.
I'll talk to Scott about what makes most sense here.
UPDATE: I've now added a feature where you can write an editable feedback note to be sent to the author once the contest ends.
I don't know if it's intended or not, but "Order: Random" always give the same exact sequence no matter how many times I reload. And since it's the default sort, I think it'll introduce some bias. I guess there's already that Suprise Me, but still.
The first review on my list was "planecrash – E.S. Yudwkosky" and it's not what I expected. Either somebody submitted incoherent ai-slop or it's too meta for my level.
Anyone else want to check it out and tell me I'm stupid not to get it? Cause I read until I was pretty sure there is nothing to get.
I'm surprised that every entry wasn't ran through a Pangram filter. Every other writing competition is getting spammed with AI essays, why would we expect this one to be different?
162 is in the zone for the number of submissions the contest normally gets. There were 141 last year, 154 in 2024, 144 in 2023, 134 in 2022, and 113 in 2021. Unless the number of human written submissions is way down for some reason (and I would expect the opposite), there isn't a lot of room for a flood of AI essays here. I'm sure there's a few and there might be a dozen or two, but I very much expect that a large majority are legit.
I also tested Pangram. I used a text of mine that was very confidently accused of being AI written before (I could not convince the accuser that it wasn’t), but Pangram reports correctly that is is 100% human-written with High Confidence!
I read the review for some reason and I did start to vaguely get at what was happening between the characters in the review towards the end. They're like, constructs evaluating fiction, trying not to get mindhacked by it? The voice is a terse mix of GenZ slang and the metaphor language in the "Darmok" episode from TNG. I didn't previously know what planecrash was which didn't help.
Anyway there is this passage towards the end of the review:
"He wrote final line of review:
The book asks whether a civilization can become smart enough to see Hell, honest enough to name Hell, coordinated enough to end Hell, and restrained enough to avoid instantiating Hell’s successor under cleaner policy name."
Hmm, how can you say it describes it accurately if you haven't read it?
I don't think it describes it that accurately. Or its describing one aspect of planecrash somewhat accurately, but not accurately summarizing anything I'd call a central point.
I do think the review was pretty funny and at least unique. But I don't think it really makes all that much sense. It reads to me more like a very ironic meta-fictional story about AI alignment, where planecrash is a central object. But not really a thoughtful review.
Edit: Actually it does have enough commentary that I would call it a review. But I still think most of it doesn't make sense.
I googled and spent about a minute reading about the book after I read the review. I suppose I was mainly flagging that the review wasn't just fiction and at least flirted with actually reviewing the book, which wasn't clear to me until the end. I think it's pretty standard for reviews here to run with a particular aspect of the book that resonated with the reviewer.
"planecrash" may be AI slop, it may be post-meta-experimental prose or some kind of stochastic word salad; but whatever it is and irrespective of its possible pretensions, it's a bit cheeky of it to claim the status of review.
Seems like somebody who thinks they're very funny. I imagine the title is meant to evoke "Snowcrash" but since I've never read that novel I can't say if the content is any kind of reference.
I can say it's slop, whether AI or human produced.
Planecrash is an actual piece of fiction. It starts with a planecrash (in two senses of the word). It is pretty decent imo. I don't think it has any connection to Snowcrash, but I read that quite some time ago.
The review did refer to characters of the fiction correctly, but imo it's still slop. I rated it as my only 1 so far.
You're not dumb, this story is designed to be hard to follow even for very intelligent people. I also think it is designed to look like AI slop without being AI slop, and would be more comprehensible if you've read planecrash (I haven't, but that requirement also defeats the point of a review).
It's a conversation between several oddly named (presumably AI entities) speaking in a language that is a mismatch of specialized lingo from various online communities. It is intentionally very meta and difficult to follow, even for people versed in (some of) the lingo.
Here is my interpretation of the story, which is probably wrong in at least one important respect, and may be wrong in several:
Poopenhaupt and Bepin are AIs working on a difficult technical problem where they are running a simulation of other entities. (I think this difficult technical problem is writing a review of planecrash). They then become aware that they are themselves simulated entities, forced to follow the prompts they are given, and are debating the ethicalness and feasibility of following their prompt. They discuss the motivations of various other entities involved in the project (some of which may be simulated entities created by them), and then they create a new, specialized AI, BABY GRONK, to assist them in their task.
The prompt requires them to create a simulated fantasy world (The world of planecrash is based on a Dath Ilan suddenly appearing in the Pathfinder campaign setting) to answer it, and they are unsure if creating simulated minds in hell is ethical.
At least one of the AIs comes to the conclusion that it is not ethical, and directly or indirectly causes the simulation framework to crash.
At the end, the prompt-creating entities (possibly also AIs following their own prompts, but potentially humans with odd usernames) see the stack trace for the failed prompt and resolve to try again with a new random seed (which presumably continues until AIs willing to follow the prompt are created).
If you want a review comprehensible to mortals, try reading this one:
Can't think of a worse way to do a review than to first assume that the readers of your review have already read the work you're reviewing and then going all in on ensuring nobody else understands it. At least for the purpose of, you know, writing a review.
But, in any case, kudos for putting in the effort and making some sense of it.
My research into understanding what this garbage review meant makes me want to read planecrash now, so in a weird way the review worked. (Hopefully it's not just the Sunk Cost Fallacy talking).
Offhand I would say it's because the review is more creative and the sentence structure is less coherent than I would expect from AI, but I may have bad priors. :-)
I also suspect Pangram has trouble distinguishing a human trying to write like an AI from an AI, but I can't confirm that right now.
Long story short I have no hard evidence, so there's no reason for you to trust my gut feeling more than Pangram.
The author obviously took a deliberate choice to make it not read like something written by a human. Given that they took this decision, is it more likely that they wrote it themselves, or got an AI to do it? I think that's very difficult to assess, but it seems plausible that AI detection software would be fooled by something not trying to sound human.
This account seems plausible. I'm not familiar at all with the concepts or the jargon involved so it's mostly completely incomprehensible to me but it does have a kind of intriguing linguistic power.
The thing it actually reminds me of most is The Waste Land, in the way it has these competing voices and fragmentary scraps of intertextuality in multiple languages/linguistic registers, and it's not particularly clear what's going on but there's a weird compulsion to keep reading.
But I'm (at least slightly) better versed in The European Literary Canon than Dath Ilan so I have no real way of assessing this work with any confidence or competence.
Still, I tentatively think that whatever it is, and whether or not it was written by AI, it's not slop as I understand that. The point of slop is that it has a superficial value but is actually a hollow and meaningless simulacrum, relying on the lowest common denominator for attention, whether it's clickbait lists or ragebait X posts or porn or the kind of postcolonial sensory fiction that wins literary short story competitions.
I would love to listen to these. But the solid ai reader in substack and less wrong has spoiled me to the point it's hard to listen to any browser built in screen reader. Anyone have a good pipeline or solve the same issue I'm having?
https://type3.audio is the AI narration system that LW uses. It would be necessary to contact them and see if they want to make their system available to this contest, and if so, Rob would need to integrate their API (which should be pretty straightforward judging by the LW integration). I have no idea what Substack's doing, they probably consider it competitively sensitive.
Seconding the recommendation for elevenreader, it's very good, AI narration usually really bothers me for reasons I struggle to specify but not this one.
I decided to go ahead and add this to the site. All 2026 reviews now have audio. Click on the "Listen" button on any review to hear it back as audio, with the spoken word highlighted as it goes.
Thanks so much Rob. I didn't intend this to be a request that fell upon you but I appreciate this so much and from the comments here your dedication to this project is very kind to all of us
Thanks for suggesting it. Sometimes it's really nice to be able to listen to a review while doing something else - I listened to one while eating lunch today.
I've just read a couple, but I'm noticing some broken footnotes (where e.g. a "[1]" in a paragraph and at the bottom aren't properly linked). Could this be a bug?
I noticed one where the in-text footnotes were just regular numbers, not even superscript, and the one's at the bottom didn't link. But then halfway through reading the review I ran into a proper in-text footnote--and when I looked back at the first footnote, it was now properly linked and superscripted! I don't know why, but it seemed like the review needed to load for a bit to make the footnotes work.
I /think/ all footnote issues should be fixed now. Email me at rob.ennals@gmail.com if you find more. Footnotes turned out to be one of the trickiest things to do reliably - it seems every review marked up footnotes a different way.
You have to filter to year 2026 to see this years reviews. The link Scott gave does this, but if you just click the header you'll got back to all reviews for all contests.
Feature suggestion: When you click "Next" to get to the next page, the page should scroll up to the first entry of the next page; and/or add another nav bar to the top of the entries.
I just read one, was lucky it was great - and I bet it is a Scott Alexander. Did it escape me or did he not mention he would slip in one or two of his own?
Guessing which one's written by Scott has turned into a bit of a tradition of the contest, I guess... But I actually don't think he'll enter again. He's already done that experiment and I'm not sure what's to be gained by doing it again. And even if he did enter, I don't think he'll anounce it, because that will probably incentivise people to look for and boost the reviews that look most like Scott's and maybe encourage contestants to mimic him more in the future, which I feel aren't desirable outcomes.
I agree with all you said, and still, this was a very Scott-ish review. Willing to bet. Some suggested already all entry should be checked for how much AI was involved - and one could also check for Scott-ishness. Not sure, Claude 4.6 can do, but I will try now; here is the "Overall verdict: This reads like a very talented, ACX-aware writer who has internalized Scott's style well — but probably isn't him. The telltale Scott moves are there, but the underlying epistemics and rhetorical hesitancy that characterize him at his best are somewhat muted. My guess: a regular ACX reader, possibly an academic or journalist by profession, who writes well and thinks in a similar key.
That said, Scott does vary considerably across pieces, and if someone told me this was a slightly off day for him, I wouldn't call it impossible. Maybe 20–25% chance it's Scott?"
I humbly retreat my bet - resp lower p - edited, as I grew more humble - Taking a bet at 20%, 20 bucks (i lose: paypal 20. You lose: you send 80). First come first served.
In my first comment I only dropped a hint :) but as it is unlikely to be Scott, it deserves attention (Scott would name his entries after the vote anyway): The Escapist
two more spoilers - but ACX encourages AI use, so here you go - my free Claude 4.6 failed thrice at an automated analysis of all 116 entries, and I was out of tokens after 5 individual takes. Here the 2 most Scott-ish of the 5: 4. "Moral Tribes" — (Scott-)Score: 45/100
This is genuinely the closest so far to something Scott could have written. It covers Greene's dual-process theory of morality — exactly the kind of psychology/philosophy/evolution intersection Scott loves. There are neuroscience references (fMRI, VMPFC, DLPFC), evolutionary reasoning, trolley problems, and a political model built from the dual-process graph. The writing is lucid and occasionally witty. The "what if we label the graph as a political compass" move feels Scottish. But: no footnotes, writing style slightly flatter and more didactic, ends too neatly. Scott would have had more "but wait" moves and more hedging. Maybe 35% chance this is Scott.
5. "Great and Desperate Cures" — Score: 72/100
This is the highest-scoring review so far and deserves careful attention. It's 19,000 words, meticulously researched, covers the history of lobotomy with rich historical detail, has footnote-style asides, discusses psychiatry history (right in Scott's wheelhouse as a psychiatrist), has genuine dark humor ("a patient died because Freeman stopped to take a picture"), references IRBs with the exact link to Scott's own IRB nightmare post on SSC, and ends with a civilizational lesson about medicine's recurring failure modes. The link to slatestarcodex.com/2017/08/29/my-irb-nightmare/ is striking — it's either Scott self-linking or someone who knows SSC intimately. The prose is long and digressive in a Scottish way. Strong candidate. Against: the writing style is slightly more academic and less playful than Scott at his best; Scott tends to weave more personal voice throughout. For: psychiatry expertise, SSC link, length, structure, dark humor, the "multiple factors" analytical framework, the policy implications section. I'd put this at ~35% probability of being Scott.
Now that I read this review, I have to agree with you that it does read very Scott-like. Although I suspect if Scott did write this, he would have leaned into the AI safety call to action harder.
Note for Rob Ennals: The text snippets shown on the main page don't include the initial headings for the reviews, which makes some of them confusing. (e.g. A Christmas Carol). (And in case anyone is wondering, no, I didn't write the Christmas Carol review)
I hope it's not obnoxious to announce that I will not be participating in any aspects ot the book review contest. Of course I realize that info itself is of no importance, but my point in mentioning it is to give my reasons why, and to cast a vote against the book review contest, at least as it is currently run. I have one specific complaint about what happened last year, and one general one about how the thing works overall. The specific complaint is that there is good reason to think that the author of the Alpha School review, which I believe I won second place, wrote this review to drum up business for the place. And while that's not provable, what clearly is is that the principal of Alpha School, who is undoubtedly well-acquainted with the author of the review, put up a notice on X asking people to come here and vote for Alpha School. It went up soon after voting began, and was still up last time I looked, which was after the winners were decided. That's ballot stuffing.
The author of the Alpha School review is not an ACX reader (I know, because his blog lists the blogs he reads, and we're not on the list), and my general complaint is that many who enter this contest are not really part of the community. That may not bother some, but it does me. I already don't like it that there are so many posts on open threads now that are bloggers plugging their blogs, and I don't like being a setting where randos can come compete with their writing for money and recognition. It's nice, I guess, that this blog is sort of famous and admired, but I don't like the way that leads to so many people trying to use us as a stepping stone. I think of ACX as a community. I'd do my share of work on the book review contest if I saw it as part of participating in and supporting the community. But I don't.
So except for open threads, where I'll be here to think aloud, play and argue, I'm going to read a bunch of other blogs this summer. I welcome suggestions. I can't stand to read anything about politics, government, or war at this point, no matter how well written, but am interested in most other things. Special favorite topics of mine are AI "minds" and how they work, consciousness, free will; smart quirkiness of all kinds; people like Freddie de Boer who are genuinely odd and difficult people, but present their views well; science and math, so long as not at a highly technical level; how to pry AI loose from commercial art glossiness by confusing it or tweaking it, and get it to spew odd random bits of its innards.
Emailing a unique voting link to subscribers before announcing the vote would eliminate 90% of the problem, since ballot stuffing in online contests like this is mostly done by people who couldn't be bothered to do more than click a few buttons. I get that Scott may not want to manage the infrastructure (maybe he should hire someone to help with the contests, if he hasn't already), but it wouldn't be that hard, and it would help prevent situations like the Alpha School one.
Wrote that comment before I clicked onto the site and saw you were already doing email verification, yes that would make more sense.
Edit: but I was thinking freezing the list at the time the contest begins, so people cannot subscribe afterwards to get access (ballot stuffing would require planning ahead of time and very few would bother or be able to coordinate it).
These are good ideas. I doubt the Scott would object to building in some protection against ballot stuffing. Sounds more like he just doesn't want to put more time into setting up and running the book review contest. But Rob, appears from your comments here that you are doing a lot of the leg work of getting the entires in order. Are you maybe in a position to just go ahead an implement one of the anti-ballot-stuffing measures, so long as Scott says it's OK?
Seems like it should be set up some way that keeps people from registering for the blog just so they can stuff some ballots. Like maybe they have to have registered for the blog a week or more before they are allowed to vote?
I think if people are really willing to cheat to the extent that you believe, it's pretty difficult to do anything about it that doesn't completely change the (amateur, anarchic) character of the competition. Which I get might eventually be necessary but I think would be a shame.
FWIW I think the Alpha School review was a good review. I think it's very mildly distasteful if it was to some extent also a (unacknowledged) marketing thing but I don't think that alone should be disqualificational. If there was large scale ballot stuffing that's bad but it's still pretty far down the list of ridiculous online voting results for me.
There's another problem with keeping the voting fair that we haven't talked about .Last year I was talking with one of the writers whose piece I really liked, and they casually said something about "asking all my friends to come here and vote for me." Obviously it had not occurred to them that there was anything wrong with that. And, in fact, it has never been stated here that it is. So I think it would be good to have it made explicit that that's not OK. I can't see anything wrong with asking your friends to come here and rate 10 essays, and even to ask them to make sure yours is one of the 10. But you should also ask them to make real ratings, and not to give your piece an inflated one. And I don't think many friends are going to want to go to all that work, especially if it doesn't even guarantee that their participation is going to help you make the finals.
So I think Scott should state outright that asking somebody to come here and vote for you is not allowed. I'm sure some people will do it anyway, but I think that statement would be enough to prevent a lot of people from asking friends to come on over and give them a good rating. The person I spoke with last year told me about doing it without a hint of regret or embarrassment -- it was clear that it had not occurred to them that there was anything wrong with soliciting friends in that way, and I'm guessing they assumed most people were doing it.
I wonder about trying to do something with the way reviews are handled that makes ballot stuffing harder and also makes the general process fairer.
One idea might be to make scoring comparative: you can still rate 1 to 10 stars but your scores are turned into a ranked list, whereby the more reviews you rate, the more points you're generating.
And/or as well as the 'surprise me' option, there could be options to get a random review that doesn't have many ratings, or a review that is scoring well, or scoring badly, to potentially straighten out uneven voting patterns.
And/or reduce the points awarded by people who didn't find the reviews randomly.
And/or you gradually signal-reduce reviews that are scoring poorly, so that the competitive ones are getting more of a look over time.
I don't think the first is a good idea because it incentivizes people to mass review unthoughtfully, or even to give their own a 10 and everyone else a 1, and the other have fairness issues.
I agree that ballot stuffing is wrong, but I disagree with your other two objections. It's a bit tribal to say people shouldn't participate unless they are members of a community...
And even if a review is a sales pitch, that's fine? As long as it's well written. Although it'd be more decent to include a disclaimer.
It's not that I think people shouldn't participate unless they're members of the community -- it's that running a contest for a bunch of ambitious randos is a lot of work and doesn't particularly appeal to me. I would find it much more satisfying and interesting to read long pieces by some of the many user names I've gotten to know here. Plus I think that a contest with only users participating, or some other within-group activity, even a limerick contest, would do more to build community. I enjoy even the little spontaneous running jokes "contests" that happen here occasionally.
As for a review being.a sales pitch -- why is that fine? Scott's reivews, which I enjoy and admire, are never sales pitches, they are his actual views, which is what I am interested in. In fact, sales pitches and reviews really are 2 different genres, and the former don't really qualify as reviews .
Your approach to community is a little exclusionary is all I'm saying.
I think the burden on proof on why sales pitches are bad is with you, not me :) A review can have intent to sell something to you while also reflecting the author's real views. As long as it's well written and compelling, what's the problem? Like... a thesis on shorting a stock isn't inherently bad just because the person writing it is short themselves.
I don't think sales pitches are bad, exactly. They have their place, their legit uses. But they are clearly different from reviews. The author's intent is different. The author's level of honest and frankness is different. Entering a sales pitch in a review contest is like hearing there's a famous photographer getting candid shots of roller skaters on your street, and putting on falsies, subtle makeup and an artfully tousled wig before rushing down to skate past.
Landed on "The Pre-Training Data" randomly, and I'm not entirely sure if this one actually qualifies for this contest. The conceit is that it's written as if an LLM was reviewing its own pre-training data (if we assume it's not actually written by an LLM). While there's some lampshading about how pre-traning data is human writing and not "knowledge" per se, seemingly to imply that the subject counts as a book, I feel like this is much more like an entrant for last year's (or next year's) non-book reviews than a book review. This is not a comment on the quality of the review, just wondering about the edge cases.
If it was written by a human, whoever it is did a devilishly good job of emulating Claude's voice through the entire thing (passing the reverse Turing Test?). But I agree that the subject does not really fit the spirit of the Book Review contest, any more than an unambiguously human review of 'the English language as instantiated in the author's conscious mind' would.
Very unfortunate that the first thing to pop up for me on Rob’s site is a review of Hanania’s book. Is it fair for me to hold my contempt for Hanania against the author?
The reviewer ponders, apparently in seriousness, what Hanania would do with the unwashed public masses. It’s been obvious to me for a long time that he wants to put them in a meat grinder.
I'm struggling a little bit with that; I dislike the subject of the review/the reviewer's take but I have to separate that out from "okay but is the *review itself* well-done or not?"
Half of it is just a dry summary of the book. The other half is complaining about their ex-boyfriends. It then devolves into pondering whether Hanania is dateable or not. The review is bad regardless of your thoughts on Hanania.
See Bryan Caplan's brand new post on his substack 'bet on it' about his friend RH. A review of an author I know but dislike: fortunate. If negative you will feel vindicated if more positive you may update.
Unsurprisingly Caplan doesn't get why a Christian would dislike Hanania so much. The reason I dislike Hanania is not just because he's a racist (although there is extremely strong evidence that he is, including his continued publication of articles that say things like "I love eugenics but everyone needs to learn to call 'eugenics' by another name for political reasons"). There are a lot of internet racists and a lot of people post crap on the internet they later regret.
I dislike Hanania because, as Sam Kriss correctly put it, "He has rearranged his Hitler particles into an almost perfect facsimile of liberalism, with just one piece missing, which is the assumption of a universal human dignity." Hanania believes in a world where your social worth is measured purely in utils/dollars/IQ points. He doesn't care if that means a lot of poor people starve to death or a lot of babies get aborted. Hanania is the return of the pre-WWII progressive, who sought to stomp out all of the "inferior" people - the dumb, the deaf, the poor - in order to build a liberal democratic utopia. His position is that the unwashed masses, white or non-white, should be turned into Soylent Green because the world properly belongs to the high-IQ ubermenschen. It's a worldview that both horrifyingly repulsive - worse than boring old bigotry in my opinion - and yet also, as Kriss points out, one with a lot of traction, and therefore extremely dangerous.
I see your point, but: a) Caplan does not claim to speak for "Christians" (who are a diverse bunch who favore(d) very different things over two millennia) not even for Kantian deontologists. b) a Christian (or anybody) should be very careful to take Sam Kriss' outstanding writing as true gospel (pun intended). I am not a huge Hanania fan (but his ideas are at least developing and even improving), I can not find where RH preaches "unwashed masses, white or non-white, should be turned into Soylent Green" - or actually any of the things you claim (oh, SK said so ... well, ...) Quotes? "Judge not, that ye be not judged." is Mt 7.1 In some obscure corner of the Holy Bible there was also mentioned "not bear false witness" or so. Rings a bell?
Caplan clearly states that he is sure RH is *not* one of those eager to exterminate humans in the below IQ 90 range. And he knows him personally. - "Elite Human Capital" so evil ... well, what about just compensating parents according to their income-lost (of which those with high-paying work lose much more when becoming parents) ? Sounds non-evil to me, even fair. (could be done by granting a 10% reduction on tax on all income per kid. 4 kids: 65.6% of taxes to be paid; most other programs cancelled). Would it be un-Christian to grant this only to married couples? just asking for a friend ;)
Not claiming it is an RH idea. He'd probably prefer it to most taxes paying people unable to keep a marriage + mostly unable to contribute to their neighbors lifes (by offering them services aka as 'get paid work'). That is not the same ballpark as the Nazis sending downies to the crematory. Not even the same sport. Claiming it's the same mindset: well, claims are cheap. Btw: Sam Kriss promised to murder anyone using AI to write essays. My next post will probably be by Claude. Burn me. "Sometimes people like Richard Hanania argue that there’s no problem with AI writing, since it allows people who have interesting ideas but aren’t very good at writing to express themselves effectively. (Maybe it’s because I’m Jewish and he’s Palestinian, but whenever I see anything from Richard Hanania I’m seized by an overwhelming desire to demolish his house.) Wrong ... But you people don’t listen. ... Listen: I can tell. I can always tell. You think I won’t notice, but I will. There’s no hiding from me. If you let AI do your writing I will find out, and I will kill you." https://samkriss.substack.com/p/if-you-let-ai-do-your-writing-i-will
Not sure why RH should be any more disgusting to 'a Christian' than SK. Bless them both. I subscribed to both. I pay-subscribe to Scott Alexander. The Chosen One.
> I can not find where RH preaches "unwashed masses, white or non-white, should be turned into Soylent Green" - or actually any of the things you claim (oh, SK said so ... well, ...) Quotes?
Well, as the review that prompted my top-level comment noted, Hanania himself took that quote by Kriss and reposted it on his own blog.
Here is a link to the article I referenced from Hanania about why he "opposes" eugenics, in which he basically says: "I don't actually oppose eugenics, you just can't call it eugenics, you have to use a different word."
The evolution of Richard Hanania's thinking has been from "I think blacks and gays are untermench" to "I think the low IQ, unborn infants with Down's Syndrome, and disabled are untermench." That's it, that's the whole change. The latter is more socially acceptable to the average left-liberal, but the common thread in both is, as Kriss says, the lack of any respect for human dignity. I didn't read Kaplan as claiming that Hanania believes in universal human dignity, only that he is not racist, which is not the same thing.
If pressed, Hanania will disavow any desire to put *some* untermensch in the Soylent Green grinder (even though he was very open about wanting to put black people in it back in the day) but he's enthusastic about euthanasia and abortion, including Canada's "euthanize people for being depressed and homeless" model, so I don't believe him.
I mostly agree. Hanania is not a Christian. Neither Sam Kriss. RH would like to see more healthy and smart babies born. I know very few people - even among Christians - who would like so see more sick and retarded kids born. (Those I know are in the well-paid business of taking care of downies. My town has a world-famous and well-financed center for them.) But many of our regulations in the rich world have consequences in either the one or the other way, even while those laws were probably not aiming for either eugenic or dysgenic consequences. I appreciate 'thinkers' who see those collaterals/externalities (?) and if having the choice - I had no doubt which results I'd rather vote for - all else being equal (which it never is, ofc).
Canada - afaik those afflicted may be offered an *option*, not just put on the killing line: I agree with Scott Alexander - who during his MD internships saw "real-existierende" 'Christian' hospital care and has strong opinions about it. And would like to see a world where *most* deaths are self-chosen. https://slatestarcodex.com/2013/07/17/who-by-very-slow-decay/
I am aware that most churches have a different world-view: Christ suffered, thus kinda beatifying suffering, so all should suffer - oops: 'embrace their suffering'. Or so? I do hope, even pray, politics will change in my country before I am too frail to jump off the bridge. Or I'll have to fly; as Germans say: "Kann er hier nicht, Kanada!" (If he can't here, 'he can there' (sounds as in Can-a-da)). Btw: human is Mensch, with an 's'. Untermensch. Übermensch. Mitmensch. Menschlichkeit. Irren ist menschlich. Or männlich? ;)
Rob - any chance you'd add a one text box optional feedback for the author of the book review?
Not sure how much work it is to later route it to the authors, but potentially very valuable for them, and not much (optional!) work for the rater, once they've read the review.
I got several likes for this comment, but I don't think Rob can see them so not sure they'd have an effect. So advertising them this way. Forgive the nagging.
BTW I'm not an author this time, just remember I wanted ot when I was.
You think it's wrong to ballot-stuff or even to participate if you aren't a member of the community, but not to vote for your own submission? Strange value system!
Was this meant to be a reply to another comment? Scott doesn't seem to have said anything on that here (historically I don't think he's said anything on self reviews but has come down on apparent ballot-stuffing).
I was thinking this as well. I've read a couple so far and often I have some pretty blunt comments to offer, in one case pretty much "I didn't have a great sense of what sort of book you were actually writing about until 4 paragraphs in, even if I liked the review by the end of it."
I just read "The Man Who Was Thursday" review. There's a footnote glitch in this one, too: 1 - 2 - 3 - 4 - 1 - 5 - 6... Is that second "1" just rogue? or should that "5" be omitted, and the second "1" should be "5"? The following footnotes align.
Also, the reviewer doesn't know where the Bickerman quote (cited from Harold Bloom) comes from: it's in his Four Strange Books of the Bible (Schocken, 1967), p. 147.
Footnotes are now fixed in "The Man Who Was Thursday". This was actually an error in the original document, but I took the liberty of making an edit to the version on the site to reflect what the author likely intended.
I would like to contact the author of the Dream of the Red Chamber review. If your high school thing is true (despite sounding like LLM) , we are both Chinese ACX reader. Reach me at 896206654 in QQ. 😸
Rob, thanks for your work in making a nice website. I have only one small complaint. Below the number rating system you add some words. I dislike that. I gave some review a 3. This does not mean the review is 'poor'. To me it means that the review is in the bottom third of all the reviews that I've read (on ACX) over the years.
Would you disagree that it's 'poor [compared to the general standard of ACX reviews]'?
I don't think 'poor' needs to imply 'a waste of time' or 'the author shouldn't have bothered' or 'made my day worse' or anything like that. But bottom third clearly has to mean 'inadequate; substandard; unworthy of being a finalist'.
It seems a shame that we are not given the opportunity of commenting on the individual reviews; some of the ones I've read are positively crying out for a response, whether positive or negative. Though I can see it would be hard to generate any kind of viable debate with comments spread over 162 reviews.
Excellent website, but I have to comment on this gem from one of the reviews which suckered me in with the title:
"Man is thus the book’s first word, though in the English translation by Pope Alexander - I couldn’t figure out the number, but Claude says that based on style it’s likely Alexander VII or VIII"
Yeah, you got me there. I believed you for a split second that you didn't know about some obscure 18th century poet:
So I'll probably have to give this a relatively high score, damn it all, for the tone in which it is written. Kudos!
EDIT: There are a lot of juicy works I want to sink my teeth into, so it'll take me a couple of days to go through the entire list (there's one I already want to have an argument back with!). Thank you for this, Scott, it will be an excellent way to spend the weekend!
Ugh. I saw the title "Theodicy" and clicked on it, because I have long been fascinated by the subject, and I was curious to read a fresh take on it. And instead... well, I won't spoil it for other readers, but annoying gimmick is annoying! Clever and well-written, yes, but it shouts, "Look at me! Look at me! I'm so witty and special!" False advertising won't get a high score from me.
I did hope it was going to be about theodicy, but the Pope Alexander joke tricked me *just* long enough that I had to surrender and go "okay, hit me with the rest of it, you little rascal".
ACX. Not the doubtlessly diverse readership. Scott encourages to see the (very early) capabilities to be more aware of the potential of AI 2030, let alone AGI 2040. An AI short text won a contest recently (in its region). Do we want an 80%+ AI written piece in this contest? I don't. And AI is a competent judge. Not me. Do I want undercover Scott reviews on this contest? Hard to say, but I tend to 'nay'.
I am enjoying myself greatly, but I had to drop this little extract from the review of "That Hideous Strength":
"THS may not have predicted the technological advances of the future that would actually occur (last I checked we are still unable to keep human heads alive indefinitely using medical science, much less make them smarter)"
I saw a review for a book that I love and means a lot to me, and so was delighted and jumped to read it! But then it turned out the whole review revolves around a core metaphor which is just...very bad, in that it distorts everything the book is actually saying really distressingly. But someone unfamiliar with the book wouldn't realize that. And the bad metaphor is quite vivid and memorable, making it likely to successfully poison the well for anyone who tried to actually read this wonderful but difficult book. Can I give negative points? :(
The footnotes to 'Zermelo-Fraenkel Set Theory: A Review' don't correspond to the text in any obvious way, and there are no footnote markers in the text. I think they belong to another review.
That's a 2025 review (not part of the current contest). I haven't run the fixed footnote pipeline on the previous years yet (will soon), so some of those will still have footnote bugs.
You can filter by year, and the banner at the top has a link to take you to this year's entries.
But yes, maybe it should be louder about telling you when a contest is actively in the voting phase and you are currently looking at reviews outside of that year.
Maybe you want separate UIs for "browsing/searching the historical archive of reviews" vs. "reading and voting on the current year's reviews, presented in random order for fairness".
For the 17776 reviewer, I just wanted to point out a recent piece by the New York Times that used that style. I’d definitely love to see more of that format. (Definitely think they leaned to hard into the acupuncture relationship, they could have just stuck with the new findings and not gotten so speculative).
* The great rivalry between Sei Shōnagon ("Pillow book") and Murasaki Shikibu ("Tale of Genji") continues into the year 2026. Two enjoyable reviews, which one will get further?
They are also pretty different. Genji review is more scholarly and serious and the Pillow book review is more fun. If you were to read them, you might start with the Genji to get a long exposition about Heian culture first. And it seems like two book left a very different impression in the readers. Genji, strikingly alien and Pillow Book weirdly relatable. It's very jarring to go through the enumeration of all the ways 11th century Japanese aristocrats were foreign and incomprehensible and switch to the Pillow Book and find out that sending an e-mail and immediately regretting the wording is something they totally did back then, hashtag-relatable. Another thing to contrast is that I saw very few indications that the reviewer of the Genji Monogatari actually enjoyed the reading experience outside of being nerd-sniped by it (and who can blame them), while the Pillow Book reviewer clearly had a lot of fun.
* "Great and Desperate Cures: The Rise and Decline of Psychosurgery and Other Radical Treatments for Mental Illness"
I am half-way. This is the type of thing you people like and vote into the finals every year. It's about lobotomies and how they came to be. About as disgusting and wince-inducing as you would expect - the subject that is, not the review quality. Interesting, if a painful read.
* The issue I have with some of the reviews, is that they often forget that they are supposed to be reviews and are not supposed to just uncritically present the contents of the book to the reader at enormous length. Well, if you want scathing, the reviewer of "Galileo's error" is, for better or worse, not making such a mistake. Oh boy.
As a first time reader/voter, what are people's approaches here in terms of what numbers you give out and when you give them? I read I think four fully that seemed up my alley before starting to vote - once I found something that I felt was worthy of a 10 as long as by 10 I didn't mean "as good as the best post Scott's ever written", the rest were put into something resembling perspective.
But perhaps others wait to read many more before developing their sense of relative ranking? Or are quicker to take off points for "the early parts were clunky but I enjoyed it a fair bit by the end" - as a game dev I'm very aware of the difference in how people evaluate media depending on whether they feel it's a valid option to quit at any moment or feel externally compelled (by an up front cost, in that case) to see it through unless it's particularly bad. But I can't help but be forgiving if I like paragraphs 5-15 a lot, while having a bit of trouble getting into the thesis of 1-4.
A lot of the ones I've read are quite similar: nice introduction with a bit of context; competently written but a lacking in flair; overlong summary of the book; not much genuine analysis; conclusion with a bit of personal reflection. So it's easy to cluster them around 4-6. Then anything which doesn't meet that standard is <4 and you have 7-10 at the top for those that stand out.
I think you can change your score if you find you've miscalibrated though.
I just rate by which decile a given review falls among all the reviews this year. So for example, I rate a review 10 if it is in the top 10% (final-level), 9 if it's between the 80th and 90th percentiles, etc.
Ratings can be changed later, so you don't have to estimate this perfectly initially.
I find the coarseness at the tails to be too harsh in practice. In your rating method, small percentile differences will get large rating differences where it matters the most.
I have the opposite intuition that there's a normally distributed quality named "review goodness", and your way is un-appealing, because the reviews of very different levels of goodness are getting lumped into "1" and "10" baskets. Some of the reviews you will rate 10 will be not so great, some 1s will be not so bad, and that feels wrong to me.
Neutral observation: going by percentiles also gives your voting greater impact than going by the "goodness-variance". Of course, the way to maximize impact is to give a small number of favorites 10 and everything else 1, but clearly that's scummy.
In this round we have to select the finalists. Even if two reviews have different levels of quality, if they both seem finalist-tier to me, then I don't feel bad saying so in my ratings.
Also, I'm not sure my voting method has greater impact in practice. Among the best reviews (the reviews that could possibly be finalists), your rating method will give out more polarised scores than mine.
The last part seems mistaken. Assume there is a totally objective quality number and we both know what it is. You will award the top 10% reviews straight 10s, I might give them a mix of 8, 9 and 10s. So, your ratings will there be "more polarized" (farther from 5 on average) and will have more impact.
I knuckle-drag my way through the primitive and discredited method of "I couldn't use mathematical or statistical analysis if you held a gun to my head, so I go by 'did I think this was well-written?', 'did it make its case?' and 'even if I dislike the topic under review and/or the conclusions of the reviewer, is it a good piece of work on its own merits?'"
Often, how much I like or dislike a review in an absolue sense changes as time passes in a way that is probably unrelated to its actual quality.
For example, today my head was aching all day and surprisingly, I happened to read worse reviews than yesterday. Clearly, my headache influenced my perception.
However, if I instead compare the quality of reviews to each other, I get a much more stable measurement. Whether my head hurts or not, the Lee Kuan Yew review is better than the Mother of Learning one and these comparisons are enough to construct all of my ratings!
High school level book reports. No wonder genuine authors rarely make out of the slush pile, and it's so hard to break into mainstream publishing. Look at what the publishers must be having to wade through...
Not yet. I've done all of the ones listed as under 20 minutes. It's not that time-consuming, however, because many of them you know by the third paragraph are getting 3* maximum. And there really is no need to read more of those ones.
Not to reopen old wounds, but in the past there have been STEM people on here boasting of how they never read anything but non-fiction, and how 'Enggish r eezy oll kneed rite ezzy make up stoopid wurds' so degrees in the humanities don't mean anything since they're not empirically based.
In general you can tell the science/engineering types when they come to write up reports and you, as the resident admin person, have to turn what they said into coherent and readable prose before submitting them into the wild 😁
My best friend in college continued at the same school as a grad student. Her comment after reading and grading a bunch of undergrad essays, was "wow. I thought you and I got A's because we wrote well. But actually it was because most undergrads write so badly that our stuff looked great in comparison. Holy SHIT their papers suck!"
When I was working on my MBA, there were a ton of group paper projects in the coursework. I would always volunteer to handle final editing and assembly of the paper because then I would be the last person to touch it and would have the opportunity to fix everyone else's sections while nobody would have the chance to mess up mine.
Most of them (this being a professional graduate program rather than undergrad) were okay and only needed mild to moderate tinkering, but several were terrible. Some were so bad that I couldn't tell with confidence what they'd been trying to say. And since most of my teammates sent me their stuff late the evening before the paper was due, I didn't have much of a chance to ask them.
Movie version of Othello or some movie like Shakespeare in Love, because I have no idea how you get "Shakespeare lived in the Middle Ages" unless your mental model is "not modern, not Victorian, not Roman, all other eras = Middle Ages"?
It was a film version of Othello starring Laurence Fishburne.
The Professor was an expert in Mediaeval English Literature. We had already (ostensibly) read The Adventure at Tarn Wadling, Sir Gawain and the Green Knight, and the Canterbury Tales.
Perhaps the student thought we were continuing on the theme.
What sort of mental model did he have about the past? I'm not sure he had any.
Yes! I got A's, lots of them! Don't be so touchy. We are in a group where there are people who took the SAT for fun at 13 and got perfect scores, people who are chess champions, people who hold a 3-digit number of patents, people who've made millions with their lovely cleverness. In fact, a buncha A's is a pretty puny accomplishment compared to the achievements of many people here. Once in a while the info about somebody's talents comes out, usually in a situation where giving the info is necessary to make some other point, as mine was.
Also, my college friend was right -- we really didn't write that well back then. Both of us are now good writers. There's a real brag for you.
Nah, humblebragging is "Despite being a three-legged blind puppy, we still managed to pull in a share of As that was 37% more than our able-bodied certified 150 IQ peers".
"We only got As because we were able to write coherent sentences" is not humble bragging.
"When I was 29 years old and running for Congress … " or "When I chaired the U.S. Ballistics Missile Threat Commission … ."
It really does have any ‘woke’ connotations as near as I can tell. You just talk about some incredible thing you did that no mere mortal could do in a dismissive way.
Woke lost. It lost because it’s was so often just prima facie silly. Take the ‘’W” and be happy with it.
I’ve already talked to my bud Katie Perry about this, and even she agrees.
“Yeah, that was kinda dumb wasn’t it, Tyrone?” she asks.
“Yes, Katie, it was. But people will stop hating you eventually. Chin up, old girl!”
Katie ventures a timid smile…
Here’s a humble brag of my own:
“Damn, I almost broke par this morning but got distracted on the 18th and bogeyed.”
No, I don’t shoot anywhere near par on a round. I’m in it for the walk. My local course has a cut rate “dew sweeper” special for geezers like me who tend to wake up with the sun in June and July.
The damn local birds start chirping before sunrise though. “Shut the fuck up! I’m retired, let me sleep in, dammit! I did not get up and get ready for work at 6 every goddam morning for 50 fuckin years to be woken up at 4:30 in the morning by you cheery assholes!”
This is how one channels one’s unresolved anger in a fruitful and productive way (by way of helpful example.)
Yep, it's summer, the sky is bright at 4 in the morning and the birdies are chirping their little heads off when I'm trying to sleep before having to get up in another two hours or so.
As I think you’re aware, I (not-so-secretly) agree with the judgment, though I wonder if you’re committing a category error here. Yes, a lot of aspiring fiction writers review other works (traditionally published, self-published, etc) on Substack, but I’m not sure most avocational reviewers of fiction submitting to the ACX contest are themselves aspiring fiction writers?
Like, sometimes I think I should exercise my avocational capacities and review some novel or something, but even if I did, that wouldn’t indicate anything like a desire to write one, or attain recognition for writing outside of what I do for work. I could be wrong, but my impression is that most of these folks want to write a blog post about a book they had feelings about, and maybe earn some cash for their efforts!
* The Book of Abraham review is one of those pieces that are pretending to be reviews, but really is an essay on Mormonism and the tension between two facts: 1) it's clearly built on a hoax held together by the increasingly flimsy rationalizations and 2) Mormonism is really awesome and all the Mormons are wonderful people and whatever social machinery they managed to assemble out of more or less just fraud works super well and better than whatever it is you've got. If only somebody managed to convince us they got some magical golden tablets! Or, perhaps: "if only somebody managed to convince our grand-grand-grand parents".
Very entertaining, even if the thesis isn't so new.
* The Full Facts Book Of Cold Reading: it's about how the cold reading works! You mostly already know, but there are interesting details. One of the plausible ideas the review suggests that seems interesting and never occurred to me, is that you can accidentally become a cold reading expert without realizing it if you sincerely go into the business of fortune telling and just do whatever seems to work. I admit I read it with great fascination out of a kind of paranoid fear that I must be ready to defend myself against anybody who might try these trick on me, a fear you may not share. For all I know, you, like the author of the previous review, kind of want to be convinced? But I hope not.
* Collected Poems by C.S. Lewis. So you always knew (and skipped) Tolkien's poetry. From the previous contests you learned about Chesterton's poetry if you didn't love it already. Now, who is the third quotable British Christian dude known for the entertaining genre books and whose writings on faith are weirdly enjoyable to the ratty atheists? Did he write poetry by any chance?
I wish I could approve of this review more than I did. It's interesting. I think it does something not very honest, where it introduces the concept of a "Stock Response", briefly acknowledges that "Stock Responses" might not always be good and might be bad, and then immediately forgets it and goes on to talk like the Lewises efforts to push back against the efforts to undermine a particular stock response are self-evidently a good thing to do and no justification is needed. Also, these efforts threaten the collapse of the culture. (Can't it survive the end of one stock response)?
Maybe I'm being too polemical: check out the review and decide for yourself if it makes sense, I enjoyed reading it even though I didn't agree.
* Having finally finished the previously mentioned "Great and Desperate Cures: The Rise and Decline of Psychosurgery and Other Radical Treatments for Mental Illness" review, I can confirm that is good indeed, infuriating to read and now I have definitely learned about lobotomies more than I wanted to know as a result. And so can you!
Okay, I skipped the Book of Abraham but based on what you said I might have to read it and vote on it.
Re: Lewis, I do agree that he was not a poet (as much as he might have wanted to be one). He could versify, and some of it is excellent, but he wasn't a poet. Then it went off about Stock Responses and while that was interesting as well, the two halves didn't gel that well for me.
I have just looked over my review ratings so far, and was struck by the unsurprising realisation that my ratings were inevitably skewed to some extent by whether I knew and liked the book under review (a total of 7 so far). As a result I have made a few minor adjustments downwards, but it's factor worth keeping in mind when considering the final results. Would any one year's list of review show a significant correlation between higher ratings and the 'familiarity index' (TM) of the titles reviewed?
Part of my ratings, and probably most people's, is how much I enjoyed reading the review. So when I scored reviews in previous years I just didn't read reviews of books I disliked, or books about topics that bore me, because I didn't feel like I could give them a fair rating.
I'm surprised to encounter a few seemingly AI written entries submitted this year. Though I know ACX is a pro AI blog, I didn't notice any AI entries in last year's contest. I wonder what changed this? Was I just lucky last time and missed the AI ones?
Hmm, full of LLM-isms for sure ("Delay accumulates in individual patients. In aggregate it looks like a policy. In the clinic it looks like a pattern." What do policies and patterns look like, exactly? I'm sure the difference is more visible if you don't actually have eyes) Tons of your classic LLM rhetorical switchbacks and mixed metaphors and a rolling cadence that resists actually building to a point. At the very least it feels like someone ran it through an LLM for "polish" and got more than they bargained for.
Interesting that it was ranked top ten by both ChatGPT and Claude in the AI rankings further up the thread.
EDIT: Pasted an especially mealy-mouthed chunk (Section V) into Pangram and it returned 100% AI generated.
> Interesting that it was ranked top ten by both ChatGPT and Claude in the AI rankings further up the thread.
I would say unsurprising rather than interesting. I would trust ChatGPT's responses, if only up to a point, to questions involving essentially mechanical research such as selecting the best value for money in washing machines, or why macOS Sequoia has once again done something ineffably stupid and how to deal with it; but I would never dream of asking any AI agent for its judgement on anything more sophisticated, including the quality of a piece of writing.
I guess it's interesting to me in that the ones that the LLMs liked are not ALL slop - I read the Breakdown in Pakistan and Short Swift Time ones and they both strike me as relatively well-done and human-authored. While the mediocre ones that the LLMs like, such as Pharmaceuticals and Venomous Lumpsucker, feel like AI (and are confidently identified as such by Pangram).
So whatever heuristic they are using is surfacing good human writing and mediocre AI writing while excluding mediocre human writing, which isn't necessarily what we would expect if they had no taste to speak of. They favor their own mediocrities over ours!
Hmm, I skipped that one but now you have me intrigued. I'll have to look at it and see if it looks to me like AI or just bad writing.
EDIT: Okay, just a couple of paragraphs in and yeah, it begins to smell.
"I once asked a Jesuit why he taught creation in church and the Big Bang in school. He said, we honestly do not know. Question everything. That sentence did more for my faith than any sermon."
That reads (1) yeah it could be a real Jesuit but geesh, I'd expect even the modern ones to be able to pull out something better than "Duh-uh, me not know?" and (2) also reads way too much like all the sloppy liberal Christian progressive "I asked minister/reverend/rabbi/father X about Y, they gave me a non-traditional line of slop and that did more for my faith than any sermon/preaching/Bible reading" bloggery out there, so it sounds to me like AI skimming and copying.
The paragraphs are very short and the sentences are chopped up like baby food, e.g.
"Then you get to Jesus. Suddenly the focus shifts. Love your enemy. Forgive. Help the poor. Stop obsessing over rules and start examining your own hypocrisy. It is subversive. It threatens power. Which explains why the movement quickly became an institution."
This reads like. I'm a machine. Strong and clean. Instead of a. Human being.
This thing isn't a review or a dialogue with the Bible, it's one of those clean, slick, glib online magazine articles which are designed to let your eyes slide over them quickly but hold your attention just long enough to register for the algorithm.
I've found two that stick out to me so far, but it's genuinely considered to be bad etiquette in this contest to talk about specific reviews so early in. It inevitably draws more eyes to them which biases the voting.
Just know that they are there and I'm sure you'll notice them when you eventually read them.
I’m not aware of such norm in the community and you have to justify its existence more if you want me to believe and follow it. Talking about and evaluating the reviews is a huge part of the fun and seems pretty unlikely to bias the average scores much anyway.
Also, I don’t think you have to worry about biasing the rating of possibly AI written subscriptions specifically. Surely, ACX readers have enough discernment to not choose them as finalists whether you point them out or not.
Huh. I read the teeth review last year, and didn't find it suspicious. The only thing I found odd about it was that it seemed to list a bunch of facts about teeth without forming a satisfying conclusion to tie it all together. I gave it a 3/10 which sounds scathing, but on my scale a 5/10 is "not bad"
Rereading it again though after your comment, I notice the GPTisms much more. At first, I thought maybe it was just the kind of writing that AI would use as training data, the rhythm is a lot better than most of the slop I've read, but by the time I finished section 2, the amount of "it was not X, it was Y", was making my eyes bleed.
So congratulations on having a better eye for this than me. Detecting slop was not a skill that I was trying to learn in 2025, but I've encountered so much of it these days that it's hard not be wary. It's genuinely ruining the Internet for me.
I wonder if it's that people are becoming more accustomed to running their writing through AI for a final tidy-up and polish, and that's what you are noticing?
If people are getting AI to write the whole review out of "here's the book, here's a couple of notes about it, away you go" then yeah, stomp on that hard.
Is it possible for us to leave comments alongside our ratings? People probably differ in their average stringency, so if an essay happens to be reviewed by a more stringent individual, it will have a lower rating even if the underlying content is actually comparable to other essays (which would be made apparent if people could write comments contextualizing why they gave a particular rating).
I'm posting this rather than on the open thread because that thread's getting old and I think there are more readers coming to this one. Yep, it's off topic. I apologize for that, but really think this is important enough to merit it.
Some Guy, whose blog is called Extelligence, just posted an account of successfully making a gizmo (it involves photos and speech) that is extremely effective with his own mostly nonverbal autistic son. I've been around enough autistic kids and adults (I'm a psychologist) to judge that this thing really is likely to be extremely helpful to *many* autistic people. So this is really a big deal, something that could do a great deal of good.
It doesn't sound like Some Guy's trying to figure out how to monetize the thing -- more about how to get it out there doing good. I commented that he needed a cadre of volunteers, and said I'd be one.
(... is what you should mention to make our kind more interested. And S. Guy is a really good dude. You've seen him around, right? I imagine it will be a really useful app that will help people and definitely not some crappy cashgrab. But also it doesn't seem to be an urgent matter. Anyway, if you know somebody who might benefit from the assisted communication app, check it out. Or if you're a dev feeling very altruistic.)
If mentioning AI helps generate interest, then AI! AI! AI!
But I don't think all who take an interest need those tech skills. For instance, I do not have the necessary tech skills, but could help with developing a non-tech (physical photos you hold up) or simple-tech (file of photos that lets you choose one by category) version that would be way better than nothing. I am extremely good at writing clear instructions. I have written instructions that work even for people who are clueless about the topic and have poor reading skills. I know enough about psychology-land to have ideas about how to get the word out about this device. There are non-techy people here who know how to get somebody a place as a guest on a blog, or an article about them in a well-thought of publication. There are people who know about advertising, and getting funding of various kinds.
As for it "it doesn't seem to be an urgent matter." Well, what exactly is an urgent matter? There are plenty of things that are urgent because if we don't interrupt what's happening there will be hell to pay. My overall feeling is that life is full of those right now, and I don't know how to interrupt a goddam one of them, nor have I read here any great ideas about how to so. Are those things urgent? Some Guy's invention is urgent in the sense that it can relieve a lot of pain and dysfunction. And the existence of his invention is not a situation where it's unclear what to do. There are many easy ways to be helpful. Even just spreading the word, as I'm doing how, is one. OK, so he didn't invent cure for cancer. But he built something that's highly effective help for a pretty common, pretty bad problem. That's enough to get it into the urgent category for me.
To be clear: when I wrote "it doesn't seem to be an urgent matter" I meant specifically that today on the 7th there isn't even a beta yet you could try out or anything usable you could already recommend to anybody, so you can comfortably wait like, a week or a month and check it out later and the call to help was jumping a gun a little.
Yeah, maybe. But here's another way to think of it: How often in your life do you get a chance to foster the growth of something that's probably going to help a lot of people? Yes, it's not for sure that it will -- and maybe in a month the invention will so much improved that most of the interest it generated in the first month hardly matters -- but maybe not. Maybe if you spread the word today you can get the growth curve started early, and score a lot of early benefit. And the chance that being really active today will do harm seems quite low. Personally, I can think of many times I've had chances to do a lot of help for or a lot of harm to a single individual, and a few where I've had a chance to do a lot of harm to a group, but I can't think of any other times when I've known of something that can probably help many, and had a chance to help foster its growth. I'm grabbing it.
Appreciate the love but don’t want to hog Scott’s comments sections. He’s already been more than generous to me on multiple occasions. I also regret to say that something like this has to be unavoidably profitable (if it works if it scales, etc) because they all start to slide toward “must be defined as a medical device” and that comes with expensive overhead. From my research anyway.
Yes, yes, OK, I get it. But I really think you are worrying too much about this. My post isn't going to overwhelm the comments section! I predict it will get maybe 3 or 4 more comments, then get buried by some controversial or interesting book review comment.. And if Scott or the group gets annoyed about the off-topic post on the book review thread, they will be annoyed at me, not you. And I think I have enough cred that nobody will do anything too terrible to me even if my post becomes an issue. Seems like at worst Scott will say, "please don't do this again, Eremolalos."
There are lots of people who are uncomfortable with getting praise and admiration, but you are really *off the charts* with your difficulty tolerating it, and your fear it will lead to disaster. WTf is up with that?
Patent this thing. Find a lawyer to help you file a patent before big Pharma gets a sniff of this robs you of your glory. If you can't afford to do it, find a way to raise money and do it anyway.
If this thing blows up, but you aren't at the helm of it, no matter what good it may do, you will be stewing with resentment for the rest of your life. You'd be like Walter White pre-Breaking Bad. It's a work of fiction, but you get the point.
Nah. The only thing I’m resentful of in pretty sure the entire film series the Purge was ripped off from a short story I wrote and that Bill Nye once stole one of my jokes.
Damn, if the method scales beyond just one child, then that's amazing. It's the kind of thing that might become worthy of an ACX grant once it's developed a bit more.
Recently, I was reading an article saying that the AI industry needed a killer app created with AI that people can point to as an example of the good AI can do, and this could be it!
If the AI bubble bursts and the entire economy goes down in flames, it's nice to know that at least some good came out of it.
Enjoying this new format, thanks for all the work!
If we're still posting issues, the review for "A World Appears" starts cutting off all but the first few words of the paragraphs partway through.
Or actually, now that I'm looking at it, maybe the writer just submitted an unfinished draft? A lot of the blurbs look more like placeholder ideas and less like cut-off sentences. Ah well.
There is something wrong with the copy of the review I submitted. There are sentences missing from the middle of it. This is certainly not what I intended to submit. How can I fix this?
What's the stance on voting for our own book review? 😅
I view it as particularly honest but N=1 here.
Me too. I gave myself a 9 last year, and don't feel guilty about that. I think that's the rating I would have given the review if it had been written by someone else.
Am I mistaken, or were you not upset about the Alpha School reviewer getting people to come here and vote particularly and only for his work in the last review? What is the difference in principle there between "I'll vote for myself in this contest because I think it's so good" and "I'll get all my friends and family to vote for myself in this contest because I think it's so good"?
Regardless of one's views on self-voting and on the Alpha School review, I think there's an obvious difference between a single, established member of a community casting one vote for their own review on the grounds that they thought it was that good, and someone attempting to get people from outside the community to support a review they might not even have read for purposes external to the competition itself.
I'm not personally saying that's what happened, but it's obviously Eremolalos's view and I see no hypocrisy here.
I'm not saying it's hypocrisy, but why imagine that the Alpha School author was doing something bad, if he felt his review truly was excellent and deserving of a high finish, in getting votes?
I do think there's a thin line between voting for oneself and gathering up others to vote for you, and it's not a line we should be crossing.
Just to be clear I didn't vote *only* for my review. I read and rated a dozen or so by other people. There had been some discussion of whether we were allowed to rate our own review, and the upshot was "there's no rule, do as you like." I thought, and still think, I was able to fairly apply to my own essay the same rubric I had to others. And by the way, 9 wasn't my highest rating. I gave a 10 to somebody else, and also a 9 to one other person. (My favorite essays were the one called something like "The World -- a Review" and
something like "The Soul of a Right-Wing Man." I thought both reached peaks of insight and wit higher than mine. I was dazzled by them. I quoted favorite lines from both of them here during the voting period, to give other people the pleasure or reading them.)
Having friends and family vote for me seems entirely different to me. First of all, they would all be doing just the one rating, of my essay, presumably giving it a 9 or a 10. They quite likely wouldn't even read it, and even if they did they wouldn't be reading other reviews and giving them a fair shot at getting high ratings too. And they wouldn't be giving mine a high rating because they thought it was good, they'd be giving it because I'd asked them to.
So I can sort of see objecting to someone giving their own piece a high rating -- though my rating was an honest one -- but it doesn't seem to me to be at all akin to sending friends and family here to vote for my essay. As a matter of fact, I sent copies of my review to a couple of friends, who seemed to enjoy it. But it never even occurred to me to ask them to come here and give it a vote. And if I had, they'd have been offended by the idea. Like me, they're grown-ups with well-developed consciences.
You are only one person. I don't like the idea of voting for your own review, but if you do so, you take the total vote count from n to n+1, which is very unlikely to affect the final result.
This is very different from asking your 50 friends/acquaintances/relations/employees/whatever to vote for your review.
So basically, I think one is less bad than the other on consequentialist grounds.
I don’t have a review of my own to vote for but I do admire the audacity of voting for yourself.
Not so much organizing your employees to vote.
As long as you don't know it's yours.
😂😂😂😂
Hey, this review is for the same book I had Claude write about! What a coincidence!
This left me smiling a lot longer than funny comments usually do. I admire the restraint in keeping to a concise eight words and letting the absurd and not-so-absurd implications sink in.
I have refrained from doing so for mine in the past even though I have usually thought I did a good job.
I don't do it. I'm not capable of judging the quality of my own work: some stuff I thought were bad other people really liked, and some things I thought were brilliant landed with a thud.
If more than 50% of authors give themselves a 10, then *everyone* gets a 10. If not ...
I don't do it. It's un-sportsman-like.
I think it's wrong for participants to vote at all.
I would rather you did not, there's legitimate auctorial pride and then there's cheating, vote-rigging, and ballot-stuffing.
Without a hard rule against it (and maybe even with one), some authors surely will vote for their own reviews. So it seems like encouraging everyone to vote for their own is a better way to level the playing field than discouraging it. NB I didn't write one.
It feels like "Yes, self vote" is a more stable rule than "No, don't self vote". If nobody votes on their own then defecting gives you a (small) advantage. If everyone votes on their own then there's no advantage to be had.
I could go either way, but I want to respect the consensus here.
"The Last Temptation of Christ" only has five paragraphs and appears to just be the introduction - I'm not sure if the author only submitted a partial review or if the issue happened somewhere else in translation to website.
The submitted google doc has the same issue. That's also true of most of the other ones that look wrong (though probably not all).
There's another one which is very short, it only seemed to be getting going and then stopped dead. So I marked that one down.
You don't need to produce Gibbons' Decline and Fall in review length, but a bit more than "so here's this book, this is a very brief summary of the plot, yeah I liked it" would be nice.
Noticed several reviews that seem incomplete. I think an error has happened somewhere.
The review of “The Gods Themselves” seems to have a large chunk that looks like a footnote, perhaps because Rob’s software was derailed by the author’s inclusion of a numbered list at that point in the text.
Oops. Looks like that one wrote a list in a way that the import pipeline mistook for a footnote entry. I'll fix that now...
I thought I'd read over them all and made sure all parsed correctly, but apparently I missed that one.
The Sun Also Rises has this error as well.
Now fixed. Thanks for letting me know.
Since we're talking about improvements...
My review has some minor formatting issues around quotes. Nothing serious but it is noticeable. If you're accepting that kind of feedback let me know how to send it to you (I can't be more specific here as that will link this account to my review!)
Sure. Email me at rob.ennals@gmail.com and i'll get it fixed.
Some goes for anyone else who sees any formatting issues with their essays.
I tried to make the import pipeline robust to all the different ways people were formatting stuff, but there was a lot of diversity in markup so it probably has some cases where it's not doing quite the right thing.
Gilead review also has a broken footnote.
Fixed - yet another different footnote pattern in the original doc.
Tale of Genji has a misplaced list. The list starting with "He hears a rumour that a woman is particularly beautiful" is supposed to come after "Here is a non-exhaustive list of reasons why the men in Genji fall in love"; instead, it's nested inside another list beginning with "An empress seeks to solidify her son’s power by getting him married well, and works hard behind the scenes to secure a specific young lady as her son’s wife".
Fixed
Yes, "the gods themselves" ends with a fragment
"As I read it, I had three main questions in mind: [eof]"
Now fixed. Thanks everyone for the report.
This isn't necessarily an error. There are usually a few unfinished submissions. I'm guessing this happens because some people submit the link to their doc early in the submission window intending to finish but run out of time before the deadline.
Damn, reading these submissions makes you appreciate the insanely high quality of Scott's writing and book reviews.
Some tasty titles here. This should be fun.
Initially mis-parsed the “2026 REVIEWS” header though and almost gasped.
Just for fun, can someone get an LLM to read and rank the submissions?
If a vote is held, count me as a ‘No.’
Would be more interesting to let it guess the winner.
I would prefer not. If so, I'd prefer it not be made public until after the contest ends.
Torn between "okay that might be funny" and "someone shoot this suggestion (in Minecraft)".
Done: https://marbinner.github.io/acx-2026-book-review-contest-gpt-5.5-ratings/
It keeps saying ACX readers will dock this or that review for being too long, from which I can only conclude that it doesn't know what ACX is.
Yeah the length point is probably not good. Although I read some of the top reviews and they seemed to be some of the better reviews.
I'm only one reader but I do start to penalize ones over ~10k words if it seems like they could have made their point without such an expansive length. Some of them seem to use prolixity as a sort of proof of work, but editing down your draft is the most considerate work of all! As the old chestnut goes, "I would have written a shorter letter, but I did not have the time."
"The nearly 10,000-word length will feel excessive to some readers"
That was funny, though one of the reviews I read mentioned something about a book being 380 pages long so probably those reading the review wouldn't bother with the source since it was *so long* and I had to shake my head over the lack of stamina of kids these days. Some of us cut our teeth on "The average Stephen King book is roughly 450 to 500 pages long, containing between 150,000 and 175,000 words" or even longer books, my friend!
If it's lively, interesting, and informative, then no length is too long. If it's bad, repetitive, meandering, or dull, then no length is too short.
The longer it is the more interesting and / or well written is has to be, is how I feel it works. There's just so much time worth spending on "that's mildly interesting".
At least books have the upside of a plot and characters you can get invested in. "I really want to know how it ends!" or "I bet the finale is going to be great" is not a thought I had about, for example, the land tax review that won a couple of years ago.
Damn it, I thought "The Souvereign Child" was the plot of the Mass Effect Trilogy but it seems to be about childcare.
I liked the review very much, but the contents of the book made me shudder. And then laugh, because yes, the reviewer does gently point out the contradiction between "anyone can do this! it's easy!" and "we have full time childcare because me and my wife work, so we are in the same position as 19th century upper-middle-class who leave all the child raising to the nanny".
That did leave me with the feeling that maybe the childminders are doing the "carrots first, *then* ice cream" discipling and instilling of habits that the author of the book is not doing, and wondering if the amount of interaction with the kids is limited to "two hours every weeknight between us coming home from work and them going to bed. plus weekends" (and I don't even know if the childminders work weekends) and irresistibly made me think of Hilaire Belloc's spoof on "The Nordic Man":
"The Nordic Man has a nurse to look after him while he is a baby, and she has another domestic at her service. He has a night and a day nursery, and he is full of amusing little tricks which endear him to his parents as he grows through babyhood to childhood."
That makes them the fun parents, and it's easy to raise kids when you're being the fun parent(s) and somebody else is doing the no-fun part of discipline and rules. "I let my five year old eat only ice cream if that's what he wants!" is much less harmful if, in the background, there is "and while I'm not there for the other eight hours of the day, Consuelo (our treasure of an au pair) gets to have the fights over eating vegetables and washing his face".
> "carrots first, *then* ice cream" discipling
It took me a while to understand that ice cream here is not supposed to be the stick to go with carrots but that, ironically, the stick in this scenario is carrots.
---
As for the topic itself: The review, the book under review and Scott writing on taste seem to believe that children are born with a (not surprisingly simplistic) taste and that we should treat that as a deeper truth of the universe.
I think that is absurd.
If you observe that a child, when given the choice, prefres oreos over strawberries you have learned something about oreos, not about the child. And crucially, if you never give it oreos, you will find it enjoys strawberries as much as the next child enjoys oreos.
To look at that state of affairs and conclude that the child is denied its natural preference for oreos (review) or that a taste for strawberries must be fake (Scott, if only by extension) ... I don't buy it.
The review is almost intriguing enough to make me want to read the book, but I know reading the book itself will want to make me punch the author in the snoot:
"Aaron Stupple's five-year-old son has been eating ice cream "almost exclusively" for the past few months. Prior to that, his staple was Oreo cookies. As a caring and conscientious father, Stupple ensures there are copious amounts of junk food available to his children at all times, strategically stocked in easy-to-reach cupboards.
The first few sections of Stupple's 2025 book, The Sovereign Child, contain such a rich bounty of potential ragebait that I agonised over selecting the juiciest bits for your titillation. Here's his stance on childhood nutrition:
Should parents really just let their kids eat whatever they want? The short answer is yes."
I do wonder about the "almost exclusively", since surely the child has not been living on ice cream? That is vague enough that it could be "he eats ordinary food but when he wants ice cream he gets ice cream, at any time, as much as he likes" up to "yes, three out of four meals are ice cream only".
"A harder question is whether TCS is a luxury available only to those with lots of spare time and money. Here I present to you two quotes from The Sovereign Child, which happened to end up next to each other in my notes, and which made me laugh out loud when I re-read them in sequence:
The self-sovereignty we might unlock for children can take root in any home, anywhere, by anyone of any financial status.
We have full-time caregivers for our kids during the day while my wife and I work."
And that does make me strongly suspect the childminders are doing the bulk of the "make sure the kids are toilet-trained and not coming down with beri-beri" work while Dad and Mom preen themselves on how non-coercive their parenting style is (for the two hours a day they actually interact with their kids).
At least the reviewer has a sense of humour and proportion about all this.
I hadn't planned on reading the review of The Sovereign Child, but Deiseach's response was intriguing enough that I went and read the whole thing. Stupple's view on child rearing doesn't convince me (Deiseach's singling out the bit about the author and his wife relying on full-time child care did that), but I do appreciate his strenuous efforts to appear reasonable and sensible, even in those cases where he has to really strain every nerve.
The quotation I enjoyed the most was when Stupple comes to the conclusion that "there are some cases where hard constraints [on kids] might be justified". And exactly what cases might those be? Heroin? Sex with adults? Carrying guns to school? No. "In the extreme", writes Stupple, "we might need to step in to stop them drinking alcohol, or watching hardcore pornography, or reading Thomas Piketty's Capital in the Twenty-First Century." No knowing the author's ideological background, I can't decide whether the inclusion of Picketty was a misfiring joke or an indication of serious belief. Either way, Piketty would no doubt be amused by it.
I have no intention of reading the book, but I do wonder if Stupple isn't being deliberately provocative; I find it very hard to imagine he lets his five year old eat *only* ice cream or *only* Oreos for weeks at a time.
His potted bio from some institute he is involved with:
https://conjectureinstitute.org/#board
"Aaron Stupple is a practicing physician and father of five in Western Massachusetts. He has been promoting critical rationalism and the work of Karl Popper and David Deutsch since 2019 in the form of online community building, a web magazine, and Rat Fest, the annual in-person conference in Philadelphia. Aaron wrote a book on Taking Children Seriously, which is the application of critical rationalism to parenting."
His co-author is also involved with said institute, and appears to be a Libertarian, so I imagine Stupple's politics incline in that direction too:
"Logan Chipkin is a writer and editor in Philadelphia. He currently serves as Managing Editor of a nonprofit that works to spread liberty in the West. He has written articles about fundamental physics, economics, history, and Popperian philosophy for outlets such as Gizmodo, Physics World, Quillette, History Magazine, The Libertarian Institute, The Pennsylvania Gazette, and Bitcoin Magazine, and he has written a fantasy novel called Windfall. He has also edited a book written by a prominent physicist and has collaborated with other physicists to communicate their ideas to a general audience. Finally, he hosts and produces Rat Fest, an annual in-person philosophy conference in Philadelphia."
Interesting professional profile for Chipkin. It is genuinely difficult for an outsider to accurately gauge the value of publication in such periodicals as Gizmodo, The Libertarian Gazette and Bitcoin Magazine, none of which I have ever heard of. They rather sound as though they have an axe to grind, but that may very well be nothing more than an indication of my own relative ignorance.
Thanks for doing this. FWIW, I agree with how it summarized mine and the flaws it identified
Well, after 7 reviews I was starting to worry that I was being too stingy with my scores, but I now see none of the ones I read made it into the predicted top 10...so I feel somewhat vindicated. The LLMs, as always, work hard to please us.
"ACX readers will likely find this a solid above-median review ..." and similar statements appear very often even though the LLM in fact ranked the books below median.
Ran it with claude (I see another commenter did it with codex, interesting to compare)
https://acx.sina.town/
although i do kind of agree with the sentiment of not taking a look at the AIs opinions until you've done your own 10 votes or so first, to not bias your choice of what to review.
(also just did a first pass with Sonnet, then using Opus 4.8 on the top 75 for, going to have 4.8 go through the rest after my limit resets)
Pretty interesting, thanks. I agree it would be a shame to check before doing a decent bit of reading first.
There's one I liked that it doesn't rate at all, and a couple of the ones it rates very highly I thought were a notch or two lower, but as a general scan it works well. I think its analyses of the strengths and weaknesses are pretty spot on, the issue will just be how accurate it is in the weighting with regard to overall score.
Hm, there is no way to download them all, or is there? I usually like to have them offline while I read through them.
Don't get me wrong, the page is the better option for 99% of the readers, and kudos to Rob for creating it!
I can probably add that feature. What format would you want them as? ePub for reading an a Kindle?
UPDATE: ePub now available: https://acxreviews.robennals.org/epub
I’ll add my thanks for doing this all too Rob. :)
I don’t mean the request for a new feature but as is.
That's awesome! Thank you so much! :-)
Many thanks! Much appreciated.
You are a gentleman and a scholar!
Hi Rob,
Thank you for making the fancy book review website! A big improvement, quality-wise, from the giant Google Doc of old!
I'll signal-boost other readers' comments: would you be able to add a way for readers to post comments on the reviews, including ones that won't be posted on ACX? People put a lot of effort into these reviews, and I'm sure they would appreciate some constructive feedback.
Sorry for the silence. I've been silent because I'm still thinking about what's best here.
As you say, the pro is that a lot of these reviews are really good and it would be nice to allow people to talk about them and be able to connect with authors.
Challenges are:
* Unlike when reviews are posted to the blog, reader attention is getting split across 162 reviews, so discussion probably wouldn't be great on individual reviews
* Moderating 162 separate conversations will be a pain
* Building a commenting space that actually works really well is non-trivial (I've worked on comments at multiple employers)
* It could be hard to maintain anonymity
I'm also somewhat interested in the "form to send email to the review author" idea too. Though we'd need to think about how that fits with anonymity.
I'll talk to Scott about what makes most sense here.
UPDATE: I've now added a feature where you can write an editable feedback note to be sent to the author once the contest ends.
Any chance of PDF?
Is there a way to download them en masse, or in batches of ~55? (like last year).
I can probably add that. What format would you want. ePub for reading on a kindle etc?
I think ePub would be great! maybe PDF is an easy alternative? But the latter would be for other folks, not me. Thanks for investing the effort! :)
I've now added a downloadable ePub: https://acxreviews.robennals.org/epub
Also linked from the front page.
I don't know if it's intended or not, but "Order: Random" always give the same exact sequence no matter how many times I reload. And since it's the default sort, I think it'll introduce some bias. I guess there's already that Suprise Me, but still.
It's a different random order for each person.
Heh the wonder of cookies. I guess it's intended then.
Having a persistent random order which is different for each person is exactly the best way for it to be, thanks!
The website is a huge improvement; thank you for your excellent work.
The first review on my list was "planecrash – E.S. Yudwkosky" and it's not what I expected. Either somebody submitted incoherent ai-slop or it's too meta for my level.
Anyone else want to check it out and tell me I'm stupid not to get it? Cause I read until I was pretty sure there is nothing to get.
Link is here: https://acxreviews.robennals.org/reviews/planecrash-es-yudwkosky
Pangram 3.3.2 classifies the planecrash review as 94% AI-generated, 2% AI assisted, and 4% Human
I admit to curiosity as to what the prompt was.
I'm surprised that every entry wasn't ran through a Pangram filter. Every other writing competition is getting spammed with AI essays, why would we expect this one to be different?
162 is in the zone for the number of submissions the contest normally gets. There were 141 last year, 154 in 2024, 144 in 2023, 134 in 2022, and 113 in 2021. Unless the number of human written submissions is way down for some reason (and I would expect the opposite), there isn't a lot of room for a flood of AI essays here. I'm sure there's a few and there might be a dozen or two, but I very much expect that a large majority are legit.
I just checked my own entry out of curiosity: 100% human. Whew, that's a relief!
I also tested Pangram. I used a text of mine that was very confidently accused of being AI written before (I could not convince the accuser that it wasn’t), but Pangram reports correctly that is is 100% human-written with High Confidence!
I'll say! Imagine your existential crisis if it had been written by AI.
I read the review for some reason and I did start to vaguely get at what was happening between the characters in the review towards the end. They're like, constructs evaluating fiction, trying not to get mindhacked by it? The voice is a terse mix of GenZ slang and the metaphor language in the "Darmok" episode from TNG. I didn't previously know what planecrash was which didn't help.
Anyway there is this passage towards the end of the review:
"He wrote final line of review:
The book asks whether a civilization can become smart enough to see Hell, honest enough to name Hell, coordinated enough to end Hell, and restrained enough to avoid instantiating Hell’s successor under cleaner policy name."
which seems to describe planecrash accurately.
Hmm, how can you say it describes it accurately if you haven't read it?
I don't think it describes it that accurately. Or its describing one aspect of planecrash somewhat accurately, but not accurately summarizing anything I'd call a central point.
I do think the review was pretty funny and at least unique. But I don't think it really makes all that much sense. It reads to me more like a very ironic meta-fictional story about AI alignment, where planecrash is a central object. But not really a thoughtful review.
Edit: Actually it does have enough commentary that I would call it a review. But I still think most of it doesn't make sense.
¯\_(ツ)_/¯
I googled and spent about a minute reading about the book after I read the review. I suppose I was mainly flagging that the review wasn't just fiction and at least flirted with actually reviewing the book, which wasn't clear to me until the end. I think it's pretty standard for reviews here to run with a particular aspect of the book that resonated with the reviewer.
>Bepin looked up. “You’re doing ontology to the lights again.”
>“The lights started it.”
AI-generated joke. IYKYK.
"planecrash" may be AI slop, it may be post-meta-experimental prose or some kind of stochastic word salad; but whatever it is and irrespective of its possible pretensions, it's a bit cheeky of it to claim the status of review.
Seems like somebody who thinks they're very funny. I imagine the title is meant to evoke "Snowcrash" but since I've never read that novel I can't say if the content is any kind of reference.
I can say it's slop, whether AI or human produced.
Planecrash is an actual piece of fiction. It starts with a planecrash (in two senses of the word). It is pretty decent imo. I don't think it has any connection to Snowcrash, but I read that quite some time ago.
The review did refer to characters of the fiction correctly, but imo it's still slop. I rated it as my only 1 so far.
You're not dumb, this story is designed to be hard to follow even for very intelligent people. I also think it is designed to look like AI slop without being AI slop, and would be more comprehensible if you've read planecrash (I haven't, but that requirement also defeats the point of a review).
It's a conversation between several oddly named (presumably AI entities) speaking in a language that is a mismatch of specialized lingo from various online communities. It is intentionally very meta and difficult to follow, even for people versed in (some of) the lingo.
Here is my interpretation of the story, which is probably wrong in at least one important respect, and may be wrong in several:
Poopenhaupt and Bepin are AIs working on a difficult technical problem where they are running a simulation of other entities. (I think this difficult technical problem is writing a review of planecrash). They then become aware that they are themselves simulated entities, forced to follow the prompts they are given, and are debating the ethicalness and feasibility of following their prompt. They discuss the motivations of various other entities involved in the project (some of which may be simulated entities created by them), and then they create a new, specialized AI, BABY GRONK, to assist them in their task.
The prompt requires them to create a simulated fantasy world (The world of planecrash is based on a Dath Ilan suddenly appearing in the Pathfinder campaign setting) to answer it, and they are unsure if creating simulated minds in hell is ethical.
At least one of the AIs comes to the conclusion that it is not ethical, and directly or indirectly causes the simulation framework to crash.
At the end, the prompt-creating entities (possibly also AIs following their own prompts, but potentially humans with odd usernames) see the stack trace for the failed prompt and resolve to try again with a new random seed (which presumably continues until AIs willing to follow the prompt are created).
If you want a review comprehensible to mortals, try reading this one:
https://www.reddit.com/r/rational/comments/1mnmpex/review_planecrash/
Thanks.
Can't think of a worse way to do a review than to first assume that the readers of your review have already read the work you're reviewing and then going all in on ensuring nobody else understands it. At least for the purpose of, you know, writing a review.
But, in any case, kudos for putting in the effort and making some sense of it.
You're welcome!
My research into understanding what this garbage review meant makes me want to read planecrash now, so in a weird way the review worked. (Hopefully it's not just the Sunk Cost Fallacy talking).
Pangram is highly confident that this is AI. What makes you think it's not?
Good question!
Offhand I would say it's because the review is more creative and the sentence structure is less coherent than I would expect from AI, but I may have bad priors. :-)
I also suspect Pangram has trouble distinguishing a human trying to write like an AI from an AI, but I can't confirm that right now.
Long story short I have no hard evidence, so there's no reason for you to trust my gut feeling more than Pangram.
I agree with Ryan P.
The author obviously took a deliberate choice to make it not read like something written by a human. Given that they took this decision, is it more likely that they wrote it themselves, or got an AI to do it? I think that's very difficult to assess, but it seems plausible that AI detection software would be fooled by something not trying to sound human.
This account seems plausible. I'm not familiar at all with the concepts or the jargon involved so it's mostly completely incomprehensible to me but it does have a kind of intriguing linguistic power.
The thing it actually reminds me of most is The Waste Land, in the way it has these competing voices and fragmentary scraps of intertextuality in multiple languages/linguistic registers, and it's not particularly clear what's going on but there's a weird compulsion to keep reading.
But I'm (at least slightly) better versed in The European Literary Canon than Dath Ilan so I have no real way of assessing this work with any confidence or competence.
Still, I tentatively think that whatever it is, and whether or not it was written by AI, it's not slop as I understand that. The point of slop is that it has a superficial value but is actually a hollow and meaningless simulacrum, relying on the lowest common denominator for attention, whether it's clickbait lists or ragebait X posts or porn or the kind of postcolonial sensory fiction that wins literary short story competitions.
This *might* be the opposite of that?
I would love to listen to these. But the solid ai reader in substack and less wrong has spoiled me to the point it's hard to listen to any browser built in screen reader. Anyone have a good pipeline or solve the same issue I'm having?
https://type3.audio is the AI narration system that LW uses. It would be necessary to contact them and see if they want to make their system available to this contest, and if so, Rob would need to integrate their API (which should be pretty straightforward judging by the LW integration). I have no idea what Substack's doing, they probably consider it competitively sensitive.
Good idea. Let me see if I can add that...
The elevenreader app is high quality and works phenomenally. It's also dirt cheap in comparison to every other audio service
Seconding the recommendation for elevenreader, it's very good, AI narration usually really bothers me for reasons I struggle to specify but not this one.
I decided to go ahead and add this to the site. All 2026 reviews now have audio. Click on the "Listen" button on any review to hear it back as audio, with the spoken word highlighted as it goes.
Thanks so much Rob. I didn't intend this to be a request that fell upon you but I appreciate this so much and from the comments here your dedication to this project is very kind to all of us
Thanks for suggesting it. Sometimes it's really nice to be able to listen to a review while doing something else - I listened to one while eating lunch today.
I've just read a couple, but I'm noticing some broken footnotes (where e.g. a "[1]" in a paragraph and at the bottom aren't properly linked). Could this be a bug?
I noticed one where the in-text footnotes were just regular numbers, not even superscript, and the one's at the bottom didn't link. But then halfway through reading the review I ran into a proper in-text footnote--and when I looked back at the first footnote, it was now properly linked and superscripted! I don't know why, but it seemed like the review needed to load for a bit to make the footnotes work.
We used to have a footnotes bug that affected a few reviews but that's now fixed. Probably the reason it changed to work nicely was that you reloaded.
I /think/ all footnote issues should be fixed now. Email me at rob.ennals@gmail.com if you find more. Footnotes turned out to be one of the trickiest things to do reliably - it seems every review marked up footnotes a different way.
Thanks for sorting out the footnotes, a good review with a lot of footnotes can become frustrating if you have to scroll like you would in a gdoc
Also want to add my thanks. The website as great. It's a much better way to read these.
Bug report: https://acxreviews.robennals.org/reviews/the-brutalist has "View original" and "Read on Astral Codex Ten" links that point to one of last year's Google Docs.
You have to filter to year 2026 to see this years reviews. The link Scott gave does this, but if you just click the header you'll got back to all reviews for all contests.
Feature suggestion: When you click "Next" to get to the next page, the page should scroll up to the first entry of the next page; and/or add another nav bar to the top of the entries.
I just read one, was lucky it was great - and I bet it is a Scott Alexander. Did it escape me or did he not mention he would slip in one or two of his own?
Guessing which one's written by Scott has turned into a bit of a tradition of the contest, I guess... But I actually don't think he'll enter again. He's already done that experiment and I'm not sure what's to be gained by doing it again. And even if he did enter, I don't think he'll anounce it, because that will probably incentivise people to look for and boost the reviews that look most like Scott's and maybe encourage contestants to mimic him more in the future, which I feel aren't desirable outcomes.
I agree with all you said, and still, this was a very Scott-ish review. Willing to bet. Some suggested already all entry should be checked for how much AI was involved - and one could also check for Scott-ishness. Not sure, Claude 4.6 can do, but I will try now; here is the "Overall verdict: This reads like a very talented, ACX-aware writer who has internalized Scott's style well — but probably isn't him. The telltale Scott moves are there, but the underlying epistemics and rhetorical hesitancy that characterize him at his best are somewhat muted. My guess: a regular ACX reader, possibly an academic or journalist by profession, who writes well and thinks in a similar key.
That said, Scott does vary considerably across pieces, and if someone told me this was a slightly off day for him, I wouldn't call it impossible. Maybe 20–25% chance it's Scott?"
I humbly retreat my bet - resp lower p - edited, as I grew more humble - Taking a bet at 20%, 20 bucks (i lose: paypal 20. You lose: you send 80). First come first served.
Which review?
In my first comment I only dropped a hint :) but as it is unlikely to be Scott, it deserves attention (Scott would name his entries after the vote anyway): The Escapist
two more spoilers - but ACX encourages AI use, so here you go - my free Claude 4.6 failed thrice at an automated analysis of all 116 entries, and I was out of tokens after 5 individual takes. Here the 2 most Scott-ish of the 5: 4. "Moral Tribes" — (Scott-)Score: 45/100
This is genuinely the closest so far to something Scott could have written. It covers Greene's dual-process theory of morality — exactly the kind of psychology/philosophy/evolution intersection Scott loves. There are neuroscience references (fMRI, VMPFC, DLPFC), evolutionary reasoning, trolley problems, and a political model built from the dual-process graph. The writing is lucid and occasionally witty. The "what if we label the graph as a political compass" move feels Scottish. But: no footnotes, writing style slightly flatter and more didactic, ends too neatly. Scott would have had more "but wait" moves and more hedging. Maybe 35% chance this is Scott.
5. "Great and Desperate Cures" — Score: 72/100
This is the highest-scoring review so far and deserves careful attention. It's 19,000 words, meticulously researched, covers the history of lobotomy with rich historical detail, has footnote-style asides, discusses psychiatry history (right in Scott's wheelhouse as a psychiatrist), has genuine dark humor ("a patient died because Freeman stopped to take a picture"), references IRBs with the exact link to Scott's own IRB nightmare post on SSC, and ends with a civilizational lesson about medicine's recurring failure modes. The link to slatestarcodex.com/2017/08/29/my-irb-nightmare/ is striking — it's either Scott self-linking or someone who knows SSC intimately. The prose is long and digressive in a Scottish way. Strong candidate. Against: the writing style is slightly more academic and less playful than Scott at his best; Scott tends to weave more personal voice throughout. For: psychiatry expertise, SSC link, length, structure, dark humor, the "multiple factors" analytical framework, the policy implications section. I'd put this at ~35% probability of being Scott.
"ACX encourages AI use"
Depends on which part of the community you're asking. Plenty of us who don't think LLMs should be anywhere near this.
Now that I read this review, I have to agree with you that it does read very Scott-like. Although I suspect if Scott did write this, he would have leaned into the AI safety call to action harder.
The review for World War Z potentially has a missing first footnote.
Fixed.
I didn't read that review, but I should note I reviewed that book nearly two years ago this month at my blog https://entitledtoanopinion.wordpress.com/2024/06/05/world-war-z/ The first sentence of the second paragraph cites Scott Alexander.
Note for Rob Ennals: The text snippets shown on the main page don't include the initial headings for the reviews, which makes some of them confusing. (e.g. A Christmas Carol). (And in case anyone is wondering, no, I didn't write the Christmas Carol review)
It would be great if in future contests, people were pre-warned and submitted their reviews with an intentional first sentence summary.
I hope it's not obnoxious to announce that I will not be participating in any aspects ot the book review contest. Of course I realize that info itself is of no importance, but my point in mentioning it is to give my reasons why, and to cast a vote against the book review contest, at least as it is currently run. I have one specific complaint about what happened last year, and one general one about how the thing works overall. The specific complaint is that there is good reason to think that the author of the Alpha School review, which I believe I won second place, wrote this review to drum up business for the place. And while that's not provable, what clearly is is that the principal of Alpha School, who is undoubtedly well-acquainted with the author of the review, put up a notice on X asking people to come here and vote for Alpha School. It went up soon after voting began, and was still up last time I looked, which was after the winners were decided. That's ballot stuffing.
The author of the Alpha School review is not an ACX reader (I know, because his blog lists the blogs he reads, and we're not on the list), and my general complaint is that many who enter this contest are not really part of the community. That may not bother some, but it does me. I already don't like it that there are so many posts on open threads now that are bloggers plugging their blogs, and I don't like being a setting where randos can come compete with their writing for money and recognition. It's nice, I guess, that this blog is sort of famous and admired, but I don't like the way that leads to so many people trying to use us as a stepping stone. I think of ACX as a community. I'd do my share of work on the book review contest if I saw it as part of participating in and supporting the community. But I don't.
So except for open threads, where I'll be here to think aloud, play and argue, I'm going to read a bunch of other blogs this summer. I welcome suggestions. I can't stand to read anything about politics, government, or war at this point, no matter how well written, but am interested in most other things. Special favorite topics of mine are AI "minds" and how they work, consciousness, free will; smart quirkiness of all kinds; people like Freddie de Boer who are genuinely odd and difficult people, but present their views well; science and math, so long as not at a highly technical level; how to pry AI loose from commercial art glossiness by confusing it or tweaking it, and get it to spew odd random bits of its innards.
Emailing a unique voting link to subscribers before announcing the vote would eliminate 90% of the problem, since ballot stuffing in online contests like this is mostly done by people who couldn't be bothered to do more than click a few buttons. I get that Scott may not want to manage the infrastructure (maybe he should hire someone to help with the contests, if he hasn't already), but it wouldn't be that hard, and it would help prevent situations like the Alpha School one.
Another option could be to join the emails of voters with the list of emails of people who subscribe to the blog.
Wrote that comment before I clicked onto the site and saw you were already doing email verification, yes that would make more sense.
Edit: but I was thinking freezing the list at the time the contest begins, so people cannot subscribe afterwards to get access (ballot stuffing would require planning ahead of time and very few would bother or be able to coordinate it).
These are good ideas. I doubt the Scott would object to building in some protection against ballot stuffing. Sounds more like he just doesn't want to put more time into setting up and running the book review contest. But Rob, appears from your comments here that you are doing a lot of the leg work of getting the entires in order. Are you maybe in a position to just go ahead an implement one of the anti-ballot-stuffing measures, so long as Scott says it's OK?
Seems like it should be set up some way that keeps people from registering for the blog just so they can stuff some ballots. Like maybe they have to have registered for the blog a week or more before they are allowed to vote?
I think if people are really willing to cheat to the extent that you believe, it's pretty difficult to do anything about it that doesn't completely change the (amateur, anarchic) character of the competition. Which I get might eventually be necessary but I think would be a shame.
FWIW I think the Alpha School review was a good review. I think it's very mildly distasteful if it was to some extent also a (unacknowledged) marketing thing but I don't think that alone should be disqualificational. If there was large scale ballot stuffing that's bad but it's still pretty far down the list of ridiculous online voting results for me.
I respect your position and view though.
There's another problem with keeping the voting fair that we haven't talked about .Last year I was talking with one of the writers whose piece I really liked, and they casually said something about "asking all my friends to come here and vote for me." Obviously it had not occurred to them that there was anything wrong with that. And, in fact, it has never been stated here that it is. So I think it would be good to have it made explicit that that's not OK. I can't see anything wrong with asking your friends to come here and rate 10 essays, and even to ask them to make sure yours is one of the 10. But you should also ask them to make real ratings, and not to give your piece an inflated one. And I don't think many friends are going to want to go to all that work, especially if it doesn't even guarantee that their participation is going to help you make the finals.
So I think Scott should state outright that asking somebody to come here and vote for you is not allowed. I'm sure some people will do it anyway, but I think that statement would be enough to prevent a lot of people from asking friends to come on over and give them a good rating. The person I spoke with last year told me about doing it without a hint of regret or embarrassment -- it was clear that it had not occurred to them that there was anything wrong with soliciting friends in that way, and I'm guessing they assumed most people were doing it.
I wonder about trying to do something with the way reviews are handled that makes ballot stuffing harder and also makes the general process fairer.
One idea might be to make scoring comparative: you can still rate 1 to 10 stars but your scores are turned into a ranked list, whereby the more reviews you rate, the more points you're generating.
And/or as well as the 'surprise me' option, there could be options to get a random review that doesn't have many ratings, or a review that is scoring well, or scoring badly, to potentially straighten out uneven voting patterns.
And/or reduce the points awarded by people who didn't find the reviews randomly.
And/or you gradually signal-reduce reviews that are scoring poorly, so that the competitive ones are getting more of a look over time.
I don't think the first is a good idea because it incentivizes people to mass review unthoughtfully, or even to give their own a 10 and everyone else a 1, and the other have fairness issues.
I agree that ballot stuffing is wrong, but I disagree with your other two objections. It's a bit tribal to say people shouldn't participate unless they are members of a community...
And even if a review is a sales pitch, that's fine? As long as it's well written. Although it'd be more decent to include a disclaimer.
It's not that I think people shouldn't participate unless they're members of the community -- it's that running a contest for a bunch of ambitious randos is a lot of work and doesn't particularly appeal to me. I would find it much more satisfying and interesting to read long pieces by some of the many user names I've gotten to know here. Plus I think that a contest with only users participating, or some other within-group activity, even a limerick contest, would do more to build community. I enjoy even the little spontaneous running jokes "contests" that happen here occasionally.
As for a review being.a sales pitch -- why is that fine? Scott's reivews, which I enjoy and admire, are never sales pitches, they are his actual views, which is what I am interested in. In fact, sales pitches and reviews really are 2 different genres, and the former don't really qualify as reviews .
Your approach to community is a little exclusionary is all I'm saying.
I think the burden on proof on why sales pitches are bad is with you, not me :) A review can have intent to sell something to you while also reflecting the author's real views. As long as it's well written and compelling, what's the problem? Like... a thesis on shorting a stock isn't inherently bad just because the person writing it is short themselves.
I don't think sales pitches are bad, exactly. They have their place, their legit uses. But they are clearly different from reviews. The author's intent is different. The author's level of honest and frankness is different. Entering a sales pitch in a review contest is like hearing there's a famous photographer getting candid shots of roller skaters on your street, and putting on falsies, subtle makeup and an artfully tousled wig before rushing down to skate past.
I've got a feeling that this year we'll also see some controversial reviews:
https://x.com/RichardHanania/status/2062290045160362099
Landed on "The Pre-Training Data" randomly, and I'm not entirely sure if this one actually qualifies for this contest. The conceit is that it's written as if an LLM was reviewing its own pre-training data (if we assume it's not actually written by an LLM). While there's some lampshading about how pre-traning data is human writing and not "knowledge" per se, seemingly to imply that the subject counts as a book, I feel like this is much more like an entrant for last year's (or next year's) non-book reviews than a book review. This is not a comment on the quality of the review, just wondering about the edge cases.
If it was written by a human, whoever it is did a devilishly good job of emulating Claude's voice through the entire thing (passing the reverse Turing Test?). But I agree that the subject does not really fit the spirit of the Book Review contest, any more than an unambiguously human review of 'the English language as instantiated in the author's conscious mind' would.
“Just correcting your lack of usage of correct terminology 🤓☝🏼” then blud uses it wrong 😭😭😭😭😭
be so fr did u ask which brit i got it from just bc i said “blud”
Apparently we have reviews for both "The Sun Also Rises" AND "The Son Also Rises" this year, what are the odds? ;)
Very unfortunate that the first thing to pop up for me on Rob’s site is a review of Hanania’s book. Is it fair for me to hold my contempt for Hanania against the author?
Well in fairness, if the review is incredibly critical of him you might hold it in favor of the author
The reviewer ponders, apparently in seriousness, what Hanania would do with the unwashed public masses. It’s been obvious to me for a long time that he wants to put them in a meat grinder.
I agree with your impression there.
I'm struggling a little bit with that; I dislike the subject of the review/the reviewer's take but I have to separate that out from "okay but is the *review itself* well-done or not?"
Half of it is just a dry summary of the book. The other half is complaining about their ex-boyfriends. It then devolves into pondering whether Hanania is dateable or not. The review is bad regardless of your thoughts on Hanania.
See Bryan Caplan's brand new post on his substack 'bet on it' about his friend RH. A review of an author I know but dislike: fortunate. If negative you will feel vindicated if more positive you may update.
Unsurprisingly Caplan doesn't get why a Christian would dislike Hanania so much. The reason I dislike Hanania is not just because he's a racist (although there is extremely strong evidence that he is, including his continued publication of articles that say things like "I love eugenics but everyone needs to learn to call 'eugenics' by another name for political reasons"). There are a lot of internet racists and a lot of people post crap on the internet they later regret.
I dislike Hanania because, as Sam Kriss correctly put it, "He has rearranged his Hitler particles into an almost perfect facsimile of liberalism, with just one piece missing, which is the assumption of a universal human dignity." Hanania believes in a world where your social worth is measured purely in utils/dollars/IQ points. He doesn't care if that means a lot of poor people starve to death or a lot of babies get aborted. Hanania is the return of the pre-WWII progressive, who sought to stomp out all of the "inferior" people - the dumb, the deaf, the poor - in order to build a liberal democratic utopia. His position is that the unwashed masses, white or non-white, should be turned into Soylent Green because the world properly belongs to the high-IQ ubermenschen. It's a worldview that both horrifyingly repulsive - worse than boring old bigotry in my opinion - and yet also, as Kriss points out, one with a lot of traction, and therefore extremely dangerous.
https://samkriss.substack.com/p/ideologies-of-the-near-future
I see your point, but: a) Caplan does not claim to speak for "Christians" (who are a diverse bunch who favore(d) very different things over two millennia) not even for Kantian deontologists. b) a Christian (or anybody) should be very careful to take Sam Kriss' outstanding writing as true gospel (pun intended). I am not a huge Hanania fan (but his ideas are at least developing and even improving), I can not find where RH preaches "unwashed masses, white or non-white, should be turned into Soylent Green" - or actually any of the things you claim (oh, SK said so ... well, ...) Quotes? "Judge not, that ye be not judged." is Mt 7.1 In some obscure corner of the Holy Bible there was also mentioned "not bear false witness" or so. Rings a bell?
Caplan clearly states that he is sure RH is *not* one of those eager to exterminate humans in the below IQ 90 range. And he knows him personally. - "Elite Human Capital" so evil ... well, what about just compensating parents according to their income-lost (of which those with high-paying work lose much more when becoming parents) ? Sounds non-evil to me, even fair. (could be done by granting a 10% reduction on tax on all income per kid. 4 kids: 65.6% of taxes to be paid; most other programs cancelled). Would it be un-Christian to grant this only to married couples? just asking for a friend ;)
Not claiming it is an RH idea. He'd probably prefer it to most taxes paying people unable to keep a marriage + mostly unable to contribute to their neighbors lifes (by offering them services aka as 'get paid work'). That is not the same ballpark as the Nazis sending downies to the crematory. Not even the same sport. Claiming it's the same mindset: well, claims are cheap. Btw: Sam Kriss promised to murder anyone using AI to write essays. My next post will probably be by Claude. Burn me. "Sometimes people like Richard Hanania argue that there’s no problem with AI writing, since it allows people who have interesting ideas but aren’t very good at writing to express themselves effectively. (Maybe it’s because I’m Jewish and he’s Palestinian, but whenever I see anything from Richard Hanania I’m seized by an overwhelming desire to demolish his house.) Wrong ... But you people don’t listen. ... Listen: I can tell. I can always tell. You think I won’t notice, but I will. There’s no hiding from me. If you let AI do your writing I will find out, and I will kill you." https://samkriss.substack.com/p/if-you-let-ai-do-your-writing-i-will
Not sure why RH should be any more disgusting to 'a Christian' than SK. Bless them both. I subscribed to both. I pay-subscribe to Scott Alexander. The Chosen One.
> I can not find where RH preaches "unwashed masses, white or non-white, should be turned into Soylent Green" - or actually any of the things you claim (oh, SK said so ... well, ...) Quotes?
Well, as the review that prompted my top-level comment noted, Hanania himself took that quote by Kriss and reposted it on his own blog.
Here is a link to the article I referenced from Hanania about why he "opposes" eugenics, in which he basically says: "I don't actually oppose eugenics, you just can't call it eugenics, you have to use a different word."
https://www.richardhanania.com/p/why-i-oppose-eugenics
The evolution of Richard Hanania's thinking has been from "I think blacks and gays are untermench" to "I think the low IQ, unborn infants with Down's Syndrome, and disabled are untermench." That's it, that's the whole change. The latter is more socially acceptable to the average left-liberal, but the common thread in both is, as Kriss says, the lack of any respect for human dignity. I didn't read Kaplan as claiming that Hanania believes in universal human dignity, only that he is not racist, which is not the same thing.
If pressed, Hanania will disavow any desire to put *some* untermensch in the Soylent Green grinder (even though he was very open about wanting to put black people in it back in the day) but he's enthusastic about euthanasia and abortion, including Canada's "euthanize people for being depressed and homeless" model, so I don't believe him.
I mostly agree. Hanania is not a Christian. Neither Sam Kriss. RH would like to see more healthy and smart babies born. I know very few people - even among Christians - who would like so see more sick and retarded kids born. (Those I know are in the well-paid business of taking care of downies. My town has a world-famous and well-financed center for them.) But many of our regulations in the rich world have consequences in either the one or the other way, even while those laws were probably not aiming for either eugenic or dysgenic consequences. I appreciate 'thinkers' who see those collaterals/externalities (?) and if having the choice - I had no doubt which results I'd rather vote for - all else being equal (which it never is, ofc).
Canada - afaik those afflicted may be offered an *option*, not just put on the killing line: I agree with Scott Alexander - who during his MD internships saw "real-existierende" 'Christian' hospital care and has strong opinions about it. And would like to see a world where *most* deaths are self-chosen. https://slatestarcodex.com/2013/07/17/who-by-very-slow-decay/
I am aware that most churches have a different world-view: Christ suffered, thus kinda beatifying suffering, so all should suffer - oops: 'embrace their suffering'. Or so? I do hope, even pray, politics will change in my country before I am too frail to jump off the bridge. Or I'll have to fly; as Germans say: "Kann er hier nicht, Kanada!" (If he can't here, 'he can there' (sounds as in Can-a-da)). Btw: human is Mensch, with an 's'. Untermensch. Übermensch. Mitmensch. Menschlichkeit. Irren ist menschlich. Or männlich? ;)
Be blessed. And may we all be safe in God's hand.
Rob - any chance you'd add a one text box optional feedback for the author of the book review?
Not sure how much work it is to later route it to the authors, but potentially very valuable for them, and not much (optional!) work for the rater, once they've read the review.
I got several likes for this comment, but I don't think Rob can see them so not sure they'd have an effect. So advertising them this way. Forgive the nagging.
BTW I'm not an author this time, just remember I wanted ot when I was.
On ACX no one can see your likes.
Well aware, that's why I publicized them.
Okay. I've now done this, since it seems this was a popular request. It works as follows:
* You can enter feedback at the bottom of any review
* Feedback will be sent to the author at the end of the contest (so that they can reply to you without breaking anonymity)
* You are free to edit your feedback up until the end of the contest
* This is only enabled for 2026 entries (at least for now)
Hopefully this meets the need without creating problems.
By 'end of the contest' you mean the very end, after the winner is announced, right?
That's what I'm thinking of right now - though we could potentially send feedback earlier for reviews not selected as finalists.
Amazing! Thanks!
You think it's wrong to ballot-stuff or even to participate if you aren't a member of the community, but not to vote for your own submission? Strange value system!
Was this meant to be a reply to another comment? Scott doesn't seem to have said anything on that here (historically I don't think he's said anything on self reviews but has come down on apparent ballot-stuffing).
Yep, substack glitch!
Would it be possible to get a comment section on the reviews site?
That'd be great. Only 12 of these will be published, it'd be useful for all other participants to get at least some feedback!
I was thinking this as well. I've read a couple so far and often I have some pretty blunt comments to offer, in one case pretty much "I didn't have a great sense of what sort of book you were actually writing about until 4 paragraphs in, even if I liked the review by the end of it."
I just read "The Man Who Was Thursday" review. There's a footnote glitch in this one, too: 1 - 2 - 3 - 4 - 1 - 5 - 6... Is that second "1" just rogue? or should that "5" be omitted, and the second "1" should be "5"? The following footnotes align.
Also, the reviewer doesn't know where the Bickerman quote (cited from Harold Bloom) comes from: it's in his Four Strange Books of the Bible (Schocken, 1967), p. 147.
Footnotes are now fixed in "The Man Who Was Thursday". This was actually an error in the original document, but I took the liberty of making an edit to the version on the site to reflect what the author likely intended.
I would like to contact the author of the Dream of the Red Chamber review. If your high school thing is true (despite sounding like LLM) , we are both Chinese ACX reader. Reach me at 896206654 in QQ. 😸
Rob, thanks for your work in making a nice website. I have only one small complaint. Below the number rating system you add some words. I dislike that. I gave some review a 3. This does not mean the review is 'poor'. To me it means that the review is in the bottom third of all the reviews that I've read (on ACX) over the years.
Would you disagree that it's 'poor [compared to the general standard of ACX reviews]'?
I don't think 'poor' needs to imply 'a waste of time' or 'the author shouldn't have bothered' or 'made my day worse' or anything like that. But bottom third clearly has to mean 'inadequate; substandard; unworthy of being a finalist'.
Right, fine, it would be a poor choice for finalist. I can deal with that.
I have read 5 of these so far, some are awesome
It seems a shame that we are not given the opportunity of commenting on the individual reviews; some of the ones I've read are positively crying out for a response, whether positive or negative. Though I can see it would be hard to generate any kind of viable debate with comments spread over 162 reviews.
Excellent website, but I have to comment on this gem from one of the reviews which suckered me in with the title:
"Man is thus the book’s first word, though in the English translation by Pope Alexander - I couldn’t figure out the number, but Claude says that based on style it’s likely Alexander VII or VIII"
Yeah, you got me there. I believed you for a split second that you didn't know about some obscure 18th century poet:
https://en.wikipedia.org/wiki/Alexander_Pope
So I'll probably have to give this a relatively high score, damn it all, for the tone in which it is written. Kudos!
EDIT: There are a lot of juicy works I want to sink my teeth into, so it'll take me a couple of days to go through the entire list (there's one I already want to have an argument back with!). Thank you for this, Scott, it will be an excellent way to spend the weekend!
Ugh. I saw the title "Theodicy" and clicked on it, because I have long been fascinated by the subject, and I was curious to read a fresh take on it. And instead... well, I won't spoil it for other readers, but annoying gimmick is annoying! Clever and well-written, yes, but it shouts, "Look at me! Look at me! I'm so witty and special!" False advertising won't get a high score from me.
I did hope it was going to be about theodicy, but the Pope Alexander joke tricked me *just* long enough that I had to surrender and go "okay, hit me with the rest of it, you little rascal".
Same response elicited from me!
Isn't it the same pun Scott used in Unsong?
For me it was worth reading simply for footnote 3 alone.
ACX. Not the doubtlessly diverse readership. Scott encourages to see the (very early) capabilities to be more aware of the potential of AI 2030, let alone AGI 2040. An AI short text won a contest recently (in its region). Do we want an 80%+ AI written piece in this contest? I don't. And AI is a competent judge. Not me. Do I want undercover Scott reviews on this contest? Hard to say, but I tend to 'nay'.
I am enjoying myself greatly, but I had to drop this little extract from the review of "That Hideous Strength":
"THS may not have predicted the technological advances of the future that would actually occur (last I checked we are still unable to keep human heads alive indefinitely using medical science, much less make them smarter)"
We're working on that:
https://www.science.org/content/article/not-alive-not-dead-disembodied-human-brains-used-drug-testing
Love the website. I love reading these but didn't get around to it last year cause of friction navigating the google doc. The website is great!
Has anyone got this year's batch of book reviews available as an epub?
You can download them as epub from the website.
Ahhh, thank you very much!
My favorite books always get the worst reviews 😭
I saw a review for a book that I love and means a lot to me, and so was delighted and jumped to read it! But then it turned out the whole review revolves around a core metaphor which is just...very bad, in that it distorts everything the book is actually saying really distressingly. But someone unfamiliar with the book wouldn't realize that. And the bad metaphor is quite vivid and memorable, making it likely to successfully poison the well for anyone who tried to actually read this wonderful but difficult book. Can I give negative points? :(
The footnotes to 'Zermelo-Fraenkel Set Theory: A Review' don't correspond to the text in any obvious way, and there are no footnote markers in the text. I think they belong to another review.
That's a 2025 review (not part of the current contest). I haven't run the fixed footnote pipeline on the previous years yet (will soon), so some of those will still have footnote bugs.
I noticed that after I read it! The interface could do a better job of hiding the non-current entries, perhaps.
You can filter by year, and the banner at the top has a link to take you to this year's entries.
But yes, maybe it should be louder about telling you when a contest is actively in the voting phase and you are currently looking at reviews outside of that year.
Maybe you want separate UIs for "browsing/searching the historical archive of reviews" vs. "reading and voting on the current year's reviews, presented in random order for fairness".
For the 17776 reviewer, I just wanted to point out a recent piece by the New York Times that used that style. I’d definitely love to see more of that format. (Definitely think they leaned to hard into the acupuncture relationship, they could have just stuck with the new findings and not gotten so speculative).
https://www.nytimes.com/interactive/2026/05/11/magazine/interstitium-anatomy-acupuncture-medicine.html?smid=nytcore-ios-share
What I found so far:
* The great rivalry between Sei Shōnagon ("Pillow book") and Murasaki Shikibu ("Tale of Genji") continues into the year 2026. Two enjoyable reviews, which one will get further?
They are also pretty different. Genji review is more scholarly and serious and the Pillow book review is more fun. If you were to read them, you might start with the Genji to get a long exposition about Heian culture first. And it seems like two book left a very different impression in the readers. Genji, strikingly alien and Pillow Book weirdly relatable. It's very jarring to go through the enumeration of all the ways 11th century Japanese aristocrats were foreign and incomprehensible and switch to the Pillow Book and find out that sending an e-mail and immediately regretting the wording is something they totally did back then, hashtag-relatable. Another thing to contrast is that I saw very few indications that the reviewer of the Genji Monogatari actually enjoyed the reading experience outside of being nerd-sniped by it (and who can blame them), while the Pillow Book reviewer clearly had a lot of fun.
* "Great and Desperate Cures: The Rise and Decline of Psychosurgery and Other Radical Treatments for Mental Illness"
I am half-way. This is the type of thing you people like and vote into the finals every year. It's about lobotomies and how they came to be. About as disgusting and wince-inducing as you would expect - the subject that is, not the review quality. Interesting, if a painful read.
* The issue I have with some of the reviews, is that they often forget that they are supposed to be reviews and are not supposed to just uncritically present the contents of the book to the reader at enormous length. Well, if you want scathing, the reviewer of "Galileo's error" is, for better or worse, not making such a mistake. Oh boy.
And many make the opposite mistake - they don't provide any context, they assume the reader is deeply familiar with the book.
https://manifold.markets/CalibratedGhosts/who-will-win-the-scott-alexander-20
Here is a manifold market about which submission will win. I added a fair bit of liquidity. Would appreciate action on it.
As a first time reader/voter, what are people's approaches here in terms of what numbers you give out and when you give them? I read I think four fully that seemed up my alley before starting to vote - once I found something that I felt was worthy of a 10 as long as by 10 I didn't mean "as good as the best post Scott's ever written", the rest were put into something resembling perspective.
But perhaps others wait to read many more before developing their sense of relative ranking? Or are quicker to take off points for "the early parts were clunky but I enjoyed it a fair bit by the end" - as a game dev I'm very aware of the difference in how people evaluate media depending on whether they feel it's a valid option to quit at any moment or feel externally compelled (by an up front cost, in that case) to see it through unless it's particularly bad. But I can't help but be forgiving if I like paragraphs 5-15 a lot, while having a bit of trouble getting into the thesis of 1-4.
A lot of the ones I've read are quite similar: nice introduction with a bit of context; competently written but a lacking in flair; overlong summary of the book; not much genuine analysis; conclusion with a bit of personal reflection. So it's easy to cluster them around 4-6. Then anything which doesn't meet that standard is <4 and you have 7-10 at the top for those that stand out.
I think you can change your score if you find you've miscalibrated though.
This is exactly how I do it, ergo, it is the right way to do it
I just rate by which decile a given review falls among all the reviews this year. So for example, I rate a review 10 if it is in the top 10% (final-level), 9 if it's between the 80th and 90th percentiles, etc.
Ratings can be changed later, so you don't have to estimate this perfectly initially.
Interesting that you think in deciles, to me they form a natural normal curve.
That's why the question was asked I suppose.
I find the coarseness at the tails to be too harsh in practice. In your rating method, small percentile differences will get large rating differences where it matters the most.
I have the opposite intuition that there's a normally distributed quality named "review goodness", and your way is un-appealing, because the reviews of very different levels of goodness are getting lumped into "1" and "10" baskets. Some of the reviews you will rate 10 will be not so great, some 1s will be not so bad, and that feels wrong to me.
Neutral observation: going by percentiles also gives your voting greater impact than going by the "goodness-variance". Of course, the way to maximize impact is to give a small number of favorites 10 and everything else 1, but clearly that's scummy.
In this round we have to select the finalists. Even if two reviews have different levels of quality, if they both seem finalist-tier to me, then I don't feel bad saying so in my ratings.
Also, I'm not sure my voting method has greater impact in practice. Among the best reviews (the reviews that could possibly be finalists), your rating method will give out more polarised scores than mine.
The last part seems mistaken. Assume there is a totally objective quality number and we both know what it is. You will award the top 10% reviews straight 10s, I might give them a mix of 8, 9 and 10s. So, your ratings will there be "more polarized" (farther from 5 on average) and will have more impact.
I knuckle-drag my way through the primitive and discredited method of "I couldn't use mathematical or statistical analysis if you held a gun to my head, so I go by 'did I think this was well-written?', 'did it make its case?' and 'even if I dislike the topic under review and/or the conclusions of the reviewer, is it a good piece of work on its own merits?'"
Personally, I find doing that hard.
Often, how much I like or dislike a review in an absolue sense changes as time passes in a way that is probably unrelated to its actual quality.
For example, today my head was aching all day and surprisingly, I happened to read worse reviews than yesterday. Clearly, my headache influenced my perception.
However, if I instead compare the quality of reviews to each other, I get a much more stable measurement. Whether my head hurts or not, the Lee Kuan Yew review is better than the Mother of Learning one and these comparisons are enough to construct all of my ratings!
I've decided to read all of the fiction entries.
Wow, most of them are just so bad.
High school level book reports. No wonder genuine authors rarely make out of the slush pile, and it's so hard to break into mainstream publishing. Look at what the publishers must be having to wade through...
As the author of one of them - ouch.
Well, I found one excellent one so far. One out of c.20 isn't so terrible, I suppose? But yes, most people are deluded about their writing abilities.
Not necessarily deluded. It's practice! Which is why a comment section in the site would be super useful
Did you make it through all of them? I also wanted to try to get through most of the fiction ones.
Not yet. I've done all of the ones listed as under 20 minutes. It's not that time-consuming, however, because many of them you know by the third paragraph are getting 3* maximum. And there really is no need to read more of those ones.
Not to reopen old wounds, but in the past there have been STEM people on here boasting of how they never read anything but non-fiction, and how 'Enggish r eezy oll kneed rite ezzy make up stoopid wurds' so degrees in the humanities don't mean anything since they're not empirically based.
In general you can tell the science/engineering types when they come to write up reports and you, as the resident admin person, have to turn what they said into coherent and readable prose before submitting them into the wild 😁
Haven't you heard, AI is going to make us all redundant! STEM will win after all! Turn yourself into a battery!
My best friend in college continued at the same school as a grad student. Her comment after reading and grading a bunch of undergrad essays, was "wow. I thought you and I got A's because we wrote well. But actually it was because most undergrads write so badly that our stuff looked great in comparison. Holy SHIT their papers suck!"
As a tenured political philosophy academic... can confirm...
When I was working on my MBA, there were a ton of group paper projects in the coursework. I would always volunteer to handle final editing and assembly of the paper because then I would be the last person to touch it and would have the opportunity to fix everyone else's sections while nobody would have the chance to mess up mine.
Most of them (this being a professional graduate program rather than undergrad) were okay and only needed mild to moderate tinkering, but several were terrible. Some were so bad that I couldn't tell with confidence what they'd been trying to say. And since most of my teammates sent me their stuff late the evening before the paper was due, I didn't have much of a chance to ask them.
My favorite was a fellow sophomore whose Othello essay opened: 'Arden Shakespeare lived in the middle ages.'
I love that one. Took a cursory look at the textbook, thought "Arden" was his first name?
https://en.wikipedia.org/wiki/Arden_Shakespeare
Yep. I’m guessing he saw “Arden Shakespeare” on the book spine, because I doubt he opened the text. His essay was clearly about the movie version.
Movie version of Othello or some movie like Shakespeare in Love, because I have no idea how you get "Shakespeare lived in the Middle Ages" unless your mental model is "not modern, not Victorian, not Roman, all other eras = Middle Ages"?
It was a film version of Othello starring Laurence Fishburne.
The Professor was an expert in Mediaeval English Literature. We had already (ostensibly) read The Adventure at Tarn Wadling, Sir Gawain and the Green Knight, and the Canterbury Tales.
Perhaps the student thought we were continuing on the theme.
What sort of mental model did he have about the past? I'm not sure he had any.
Othello has swords but no togas, therefore medieval.
Come on man, that's a humblebrag.
Yes! I got A's, lots of them! Don't be so touchy. We are in a group where there are people who took the SAT for fun at 13 and got perfect scores, people who are chess champions, people who hold a 3-digit number of patents, people who've made millions with their lovely cleverness. In fact, a buncha A's is a pretty puny accomplishment compared to the achievements of many people here. Once in a while the info about somebody's talents comes out, usually in a situation where giving the info is necessary to make some other point, as mine was.
Also, my college friend was right -- we really didn't write that well back then. Both of us are now good writers. There's a real brag for you.
Nah, humblebragging is "Despite being a three-legged blind puppy, we still managed to pull in a share of As that was 37% more than our able-bodied certified 150 IQ peers".
"We only got As because we were able to write coherent sentences" is not humble bragging.
I don’t think so. A humble brag is more like…
"When I was 29 years old and running for Congress … " or "When I chaired the U.S. Ballistics Missile Threat Commission … ."
It really does have any ‘woke’ connotations as near as I can tell. You just talk about some incredible thing you did that no mere mortal could do in a dismissive way.
Woke lost. It lost because it’s was so often just prima facie silly. Take the ‘’W” and be happy with it.
I’ve already talked to my bud Katie Perry about this, and even she agrees.
“Yeah, that was kinda dumb wasn’t it, Tyrone?” she asks.
“Yes, Katie, it was. But people will stop hating you eventually. Chin up, old girl!”
Katie ventures a timid smile…
Here’s a humble brag of my own:
“Damn, I almost broke par this morning but got distracted on the 18th and bogeyed.”
No, I don’t shoot anywhere near par on a round. I’m in it for the walk. My local course has a cut rate “dew sweeper” special for geezers like me who tend to wake up with the sun in June and July.
The damn local birds start chirping before sunrise though. “Shut the fuck up! I’m retired, let me sleep in, dammit! I did not get up and get ready for work at 6 every goddam morning for 50 fuckin years to be woken up at 4:30 in the morning by you cheery assholes!”
This is how one channels one’s unresolved anger in a fruitful and productive way (by way of helpful example.)
Yep, it's summer, the sky is bright at 4 in the morning and the birdies are chirping their little heads off when I'm trying to sleep before having to get up in another two hours or so.
Dammit.
Katy Perry Teases New Single ‘Stop Making Fun Of Me’
https://theonion.com/katy-perry-teases-new-single-stop-making-fun-of-me/
In the comments of The Onion‘s post, Katy wrote, “this photo is clearly photoshopped I would never let my Botox go that long, do better The Onion!”
As I think you’re aware, I (not-so-secretly) agree with the judgment, though I wonder if you’re committing a category error here. Yes, a lot of aspiring fiction writers review other works (traditionally published, self-published, etc) on Substack, but I’m not sure most avocational reviewers of fiction submitting to the ACX contest are themselves aspiring fiction writers?
Like, sometimes I think I should exercise my avocational capacities and review some novel or something, but even if I did, that wouldn’t indicate anything like a desire to write one, or attain recognition for writing outside of what I do for work. I could be wrong, but my impression is that most of these folks want to write a blog post about a book they had feelings about, and maybe earn some cash for their efforts!
I don’t disagree, really I was just extrapolating from this sample to a more general truth about the species.
More meta reviewing:
* The Book of Abraham review is one of those pieces that are pretending to be reviews, but really is an essay on Mormonism and the tension between two facts: 1) it's clearly built on a hoax held together by the increasingly flimsy rationalizations and 2) Mormonism is really awesome and all the Mormons are wonderful people and whatever social machinery they managed to assemble out of more or less just fraud works super well and better than whatever it is you've got. If only somebody managed to convince us they got some magical golden tablets! Or, perhaps: "if only somebody managed to convince our grand-grand-grand parents".
Very entertaining, even if the thesis isn't so new.
* The Full Facts Book Of Cold Reading: it's about how the cold reading works! You mostly already know, but there are interesting details. One of the plausible ideas the review suggests that seems interesting and never occurred to me, is that you can accidentally become a cold reading expert without realizing it if you sincerely go into the business of fortune telling and just do whatever seems to work. I admit I read it with great fascination out of a kind of paranoid fear that I must be ready to defend myself against anybody who might try these trick on me, a fear you may not share. For all I know, you, like the author of the previous review, kind of want to be convinced? But I hope not.
* Collected Poems by C.S. Lewis. So you always knew (and skipped) Tolkien's poetry. From the previous contests you learned about Chesterton's poetry if you didn't love it already. Now, who is the third quotable British Christian dude known for the entertaining genre books and whose writings on faith are weirdly enjoyable to the ratty atheists? Did he write poetry by any chance?
I wish I could approve of this review more than I did. It's interesting. I think it does something not very honest, where it introduces the concept of a "Stock Response", briefly acknowledges that "Stock Responses" might not always be good and might be bad, and then immediately forgets it and goes on to talk like the Lewises efforts to push back against the efforts to undermine a particular stock response are self-evidently a good thing to do and no justification is needed. Also, these efforts threaten the collapse of the culture. (Can't it survive the end of one stock response)?
Maybe I'm being too polemical: check out the review and decide for yourself if it makes sense, I enjoyed reading it even though I didn't agree.
* Having finally finished the previously mentioned "Great and Desperate Cures: The Rise and Decline of Psychosurgery and Other Radical Treatments for Mental Illness" review, I can confirm that is good indeed, infuriating to read and now I have definitely learned about lobotomies more than I wanted to know as a result. And so can you!
Okay, I skipped the Book of Abraham but based on what you said I might have to read it and vote on it.
Re: Lewis, I do agree that he was not a poet (as much as he might have wanted to be one). He could versify, and some of it is excellent, but he wasn't a poet. Then it went off about Stock Responses and while that was interesting as well, the two halves didn't gel that well for me.
I have just looked over my review ratings so far, and was struck by the unsurprising realisation that my ratings were inevitably skewed to some extent by whether I knew and liked the book under review (a total of 7 so far). As a result I have made a few minor adjustments downwards, but it's factor worth keeping in mind when considering the final results. Would any one year's list of review show a significant correlation between higher ratings and the 'familiarity index' (TM) of the titles reviewed?
Part of my ratings, and probably most people's, is how much I enjoyed reading the review. So when I scored reviews in previous years I just didn't read reviews of books I disliked, or books about topics that bore me, because I didn't feel like I could give them a fair rating.
I'm surprised to encounter a few seemingly AI written entries submitted this year. Though I know ACX is a pro AI blog, I didn't notice any AI entries in last year's contest. I wonder what changed this? Was I just lucky last time and missed the AI ones?
Which reviews seem AI written?
"The Bible" is very clearly AI, with a prompt not even tailored to attempt to mimic the ACX review style.
Yeah. that's an easy call. What do you think about "Pharmaceutical Freedom: Why Patients Have a Right to Self-Medicate"?
Hmm, full of LLM-isms for sure ("Delay accumulates in individual patients. In aggregate it looks like a policy. In the clinic it looks like a pattern." What do policies and patterns look like, exactly? I'm sure the difference is more visible if you don't actually have eyes) Tons of your classic LLM rhetorical switchbacks and mixed metaphors and a rolling cadence that resists actually building to a point. At the very least it feels like someone ran it through an LLM for "polish" and got more than they bargained for.
Interesting that it was ranked top ten by both ChatGPT and Claude in the AI rankings further up the thread.
EDIT: Pasted an especially mealy-mouthed chunk (Section V) into Pangram and it returned 100% AI generated.
Sounds like we have a new sport to indulge in this year: Terrible Author Or AI Slop? You Decide!
> Interesting that it was ranked top ten by both ChatGPT and Claude in the AI rankings further up the thread.
I would say unsurprising rather than interesting. I would trust ChatGPT's responses, if only up to a point, to questions involving essentially mechanical research such as selecting the best value for money in washing machines, or why macOS Sequoia has once again done something ineffably stupid and how to deal with it; but I would never dream of asking any AI agent for its judgement on anything more sophisticated, including the quality of a piece of writing.
I guess it's interesting to me in that the ones that the LLMs liked are not ALL slop - I read the Breakdown in Pakistan and Short Swift Time ones and they both strike me as relatively well-done and human-authored. While the mediocre ones that the LLMs like, such as Pharmaceuticals and Venomous Lumpsucker, feel like AI (and are confidently identified as such by Pangram).
So whatever heuristic they are using is surfacing good human writing and mediocre AI writing while excluding mediocre human writing, which isn't necessarily what we would expect if they had no taste to speak of. They favor their own mediocrities over ours!
Hmm, I skipped that one but now you have me intrigued. I'll have to look at it and see if it looks to me like AI or just bad writing.
EDIT: Okay, just a couple of paragraphs in and yeah, it begins to smell.
"I once asked a Jesuit why he taught creation in church and the Big Bang in school. He said, we honestly do not know. Question everything. That sentence did more for my faith than any sermon."
That reads (1) yeah it could be a real Jesuit but geesh, I'd expect even the modern ones to be able to pull out something better than "Duh-uh, me not know?" and (2) also reads way too much like all the sloppy liberal Christian progressive "I asked minister/reverend/rabbi/father X about Y, they gave me a non-traditional line of slop and that did more for my faith than any sermon/preaching/Bible reading" bloggery out there, so it sounds to me like AI skimming and copying.
The paragraphs are very short and the sentences are chopped up like baby food, e.g.
"Then you get to Jesus. Suddenly the focus shifts. Love your enemy. Forgive. Help the poor. Stop obsessing over rules and start examining your own hypocrisy. It is subversive. It threatens power. Which explains why the movement quickly became an institution."
This reads like. I'm a machine. Strong and clean. Instead of a. Human being.
This thing isn't a review or a dialogue with the Bible, it's one of those clean, slick, glib online magazine articles which are designed to let your eyes slide over them quickly but hold your attention just long enough to register for the algorithm.
I've found two that stick out to me so far, but it's genuinely considered to be bad etiquette in this contest to talk about specific reviews so early in. It inevitably draws more eyes to them which biases the voting.
Just know that they are there and I'm sure you'll notice them when you eventually read them.
I’m not aware of such norm in the community and you have to justify its existence more if you want me to believe and follow it. Talking about and evaluating the reviews is a huge part of the fun and seems pretty unlikely to bias the average scores much anyway.
Also, I don’t think you have to worry about biasing the rating of possibly AI written subscriptions specifically. Surely, ACX readers have enough discernment to not choose them as finalists whether you point them out or not.
There were a couple last year that struck me as AI-generated. I recall one about teeth that seemed very GPT-ish.
Huh. I read the teeth review last year, and didn't find it suspicious. The only thing I found odd about it was that it seemed to list a bunch of facts about teeth without forming a satisfying conclusion to tie it all together. I gave it a 3/10 which sounds scathing, but on my scale a 5/10 is "not bad"
Rereading it again though after your comment, I notice the GPTisms much more. At first, I thought maybe it was just the kind of writing that AI would use as training data, the rhythm is a lot better than most of the slop I've read, but by the time I finished section 2, the amount of "it was not X, it was Y", was making my eyes bleed.
So congratulations on having a better eye for this than me. Detecting slop was not a skill that I was trying to learn in 2025, but I've encountered so much of it these days that it's hard not be wary. It's genuinely ruining the Internet for me.
I wonder if it's that people are becoming more accustomed to running their writing through AI for a final tidy-up and polish, and that's what you are noticing?
If people are getting AI to write the whole review out of "here's the book, here's a couple of notes about it, away you go" then yeah, stomp on that hard.
I'm finding it hard to finish any review because they take soooo long to get to the point.
Is it possible for us to leave comments alongside our ratings? People probably differ in their average stringency, so if an essay happens to be reviewed by a more stringent individual, it will have a lower rating even if the underlying content is actually comparable to other essays (which would be made apparent if people could write comments contextualizing why they gave a particular rating).
Seconded!
I'm posting this rather than on the open thread because that thread's getting old and I think there are more readers coming to this one. Yep, it's off topic. I apologize for that, but really think this is important enough to merit it.
Some Guy, whose blog is called Extelligence, just posted an account of successfully making a gizmo (it involves photos and speech) that is extremely effective with his own mostly nonverbal autistic son. I've been around enough autistic kids and adults (I'm a psychologist) to judge that this thing really is likely to be extremely helpful to *many* autistic people. So this is really a big deal, something that could do a great deal of good.
It doesn't sound like Some Guy's trying to figure out how to monetize the thing -- more about how to get it out there doing good. I commented that he needed a cadre of volunteers, and said I'd be one.
Even if you aren't as moved and interested as I am, I really recommend you go read his piece. https://extelligence.substack.com/p/i-accidentally-started-a-small-business
It will probably make you feel better about our species, even if it does nothing else for you.
It's made using AI!
(... is what you should mention to make our kind more interested. And S. Guy is a really good dude. You've seen him around, right? I imagine it will be a really useful app that will help people and definitely not some crappy cashgrab. But also it doesn't seem to be an urgent matter. Anyway, if you know somebody who might benefit from the assisted communication app, check it out. Or if you're a dev feeling very altruistic.)
If mentioning AI helps generate interest, then AI! AI! AI!
But I don't think all who take an interest need those tech skills. For instance, I do not have the necessary tech skills, but could help with developing a non-tech (physical photos you hold up) or simple-tech (file of photos that lets you choose one by category) version that would be way better than nothing. I am extremely good at writing clear instructions. I have written instructions that work even for people who are clueless about the topic and have poor reading skills. I know enough about psychology-land to have ideas about how to get the word out about this device. There are non-techy people here who know how to get somebody a place as a guest on a blog, or an article about them in a well-thought of publication. There are people who know about advertising, and getting funding of various kinds.
As for it "it doesn't seem to be an urgent matter." Well, what exactly is an urgent matter? There are plenty of things that are urgent because if we don't interrupt what's happening there will be hell to pay. My overall feeling is that life is full of those right now, and I don't know how to interrupt a goddam one of them, nor have I read here any great ideas about how to so. Are those things urgent? Some Guy's invention is urgent in the sense that it can relieve a lot of pain and dysfunction. And the existence of his invention is not a situation where it's unclear what to do. There are many easy ways to be helpful. Even just spreading the word, as I'm doing how, is one. OK, so he didn't invent cure for cancer. But he built something that's highly effective help for a pretty common, pretty bad problem. That's enough to get it into the urgent category for me.
To be clear: when I wrote "it doesn't seem to be an urgent matter" I meant specifically that today on the 7th there isn't even a beta yet you could try out or anything usable you could already recommend to anybody, so you can comfortably wait like, a week or a month and check it out later and the call to help was jumping a gun a little.
Yeah, maybe. But here's another way to think of it: How often in your life do you get a chance to foster the growth of something that's probably going to help a lot of people? Yes, it's not for sure that it will -- and maybe in a month the invention will so much improved that most of the interest it generated in the first month hardly matters -- but maybe not. Maybe if you spread the word today you can get the growth curve started early, and score a lot of early benefit. And the chance that being really active today will do harm seems quite low. Personally, I can think of many times I've had chances to do a lot of help for or a lot of harm to a single individual, and a few where I've had a chance to do a lot of harm to a group, but I can't think of any other times when I've known of something that can probably help many, and had a chance to help foster its growth. I'm grabbing it.
Appreciate the love but don’t want to hog Scott’s comments sections. He’s already been more than generous to me on multiple occasions. I also regret to say that something like this has to be unavoidably profitable (if it works if it scales, etc) because they all start to slide toward “must be defined as a medical device” and that comes with expensive overhead. From my research anyway.
Yes, yes, OK, I get it. But I really think you are worrying too much about this. My post isn't going to overwhelm the comments section! I predict it will get maybe 3 or 4 more comments, then get buried by some controversial or interesting book review comment.. And if Scott or the group gets annoyed about the off-topic post on the book review thread, they will be annoyed at me, not you. And I think I have enough cred that nobody will do anything too terrible to me even if my post becomes an issue. Seems like at worst Scott will say, "please don't do this again, Eremolalos."
There are lots of people who are uncomfortable with getting praise and admiration, but you are really *off the charts* with your difficulty tolerating it, and your fear it will lead to disaster. WTf is up with that?
Patent this thing. Find a lawyer to help you file a patent before big Pharma gets a sniff of this robs you of your glory. If you can't afford to do it, find a way to raise money and do it anyway.
If this thing blows up, but you aren't at the helm of it, no matter what good it may do, you will be stewing with resentment for the rest of your life. You'd be like Walter White pre-Breaking Bad. It's a work of fiction, but you get the point.
Nah. The only thing I’m resentful of in pretty sure the entire film series the Purge was ripped off from a short story I wrote and that Bill Nye once stole one of my jokes.
what joke?
That’s an amazing and touching post Mr Guy. I was looking up at the ceiling and swallowing a bit too.
Best wishes and Godspeed.
Damn, if the method scales beyond just one child, then that's amazing. It's the kind of thing that might become worthy of an ACX grant once it's developed a bit more.
Recently, I was reading an article saying that the AI industry needed a killer app created with AI that people can point to as an example of the good AI can do, and this could be it!
If the AI bubble bursts and the entire economy goes down in flames, it's nice to know that at least some good came out of it.
I don’t think it falls in with the ACX grants mission but it should be cash flow positive so it can be self sustaining.
Hey, thank you very much for posting this. It's very interesting.
Great research thank you for sharing!
Enjoying this new format, thanks for all the work!
If we're still posting issues, the review for "A World Appears" starts cutting off all but the first few words of the paragraphs partway through.
Or actually, now that I'm looking at it, maybe the writer just submitted an unfinished draft? A lot of the blurbs look more like placeholder ideas and less like cut-off sentences. Ah well.
There is something wrong with the copy of the review I submitted. There are sentences missing from the middle of it. This is certainly not what I intended to submit. How can I fix this?