Links For April 2026

[I haven’t independently verified each link. On average, commenters will end up spotting evidence that around two or three of the links in each links post are wrong or misleading. I correct these as I see them, and will highlight important corrections later, but I can’t guarantee I will have caught them all by the time you read this.] 1: It’s easy to draw Venn diagrams for two or three concepts, and doable for four or five, but after that it starts to get thorny. 2: Did you know: in many countries, desecrating the national flag is illegal. In some countries, desecrating foreign flags is also illegal. But in two countries - Uruguay and Denmark - it’s only legal to desecrate your own nation’s flag, but not that of a foreign nation! (because burning the home country’s flag is protest / protected speech, but burning foreign nations’ flags threatens diplomatic relations). 3: Nicholas Weininger on why governments outsource. Just as one example, “for any given amount of services you want to provide, the [non-government] nonprofit is going to be way cheaper because their employees aren’t subject to public-sector union contract rules. You can pay them a lot less, and maybe even more importantly, you can hire and fire them at will.” 4: Spencer Nitkey on the Omelas story: For me, the whole of Omelas spins around a small paragraph, directed at the reader, that introduces the infamous suffering child: “Do you believe? Do you accept the festival, the city, the joy? No? Then let me describe one more thing.” This line splits the story almost exactly in half (~1500 words into the ~3000 word story). The first 1500 words describe the mature and intellectual and euphorically prosocial joys of this perfect city…After this line, a promise to make the world sensible to the reader, the long description of the tortured child begins. Then, at the almost-end, right before the titular ones who walk away are introduced, the narrator returns to this thought: “Now do you believe them? Are they not more credible?” This sentiment spines the story, a not-so-gentle critique of our inability to imagine radical goodness without tempering it with deep horrible and inhuman trade-offs. “You don’t believe such a wonderful place could exist? It strains credulity? What if I told you about a ruined child, Atlasian in their suffering, who holds it all up? Notice how more readily you now accede to this place?”… It’s not a literary treatment of the trolley problem; it’s a critique of its reader (and of course, therefore, a critique of the society that shapes such readers). In this read…walking away from Omelas is less about rejecting an immoral system—refusing to eat while others starve—and more about rejecting an underlying ontology that renders goodness possible only if it’s enabled by subterranean suffering. 5: Russian folklore describes the witch Baba Yaga as living in a “hut with chicken legs”. This could derive from the Sami custom of building storehouses on tree trunks to protect against predators (h/t @UralicMoon) 6: Did you know: you can “talk to the chatbot trapped in Google Translate” (h/t @goremoder, @ipsumkyle, I can’t replicate this but people in the replies can) 7: Related: Kagi Translate lets you translate from any language/style/way-of-speaking to any other. For example, here’s English → LinkedIn If you hand-change the URL, you can do whatever you want. For example, if you take the link https://translate.kagi.com/?from=en&to=linkedin&text=The+quick+brown+fox+jumped+over+the+lazy+dog and replace the term linkedin with medievaldonaldtrump, you get: To calibrate your obscurity level, the AI is able to do eliezeryudkowsky but not astralcodexten. 8: Persuasion: The Unexpected Persistence Of John Rawls. This did exactly what I wanted from a John Rawls article - admit that the exact details of the way he sets up the veil of ignorance experiment are bad and everyone knows it, then explain why he’s the greatest modern philosopher of liberalism anyway. Short summary: pre-Rawls, “liberalism” was just economic laissez-faire; Rawls created large parts of the modern story about individuals cooperating to create stable institutions, and the argument that instead of choosing a correct value system the role of the state is to create a neutral space where multiple disagreeing value systems can flourish. It also (imho correctly) calls out modern “postliberals” as pre-Rawlsian rather than post-Rawlsian, in the sense of not confronting or having a good answer to Rawls. 9: Jim O’Neill, among the most good-parts-of-Silicon-Valley value-aligned people in the current administration, has been transferred from deputy director of HHS to director of the National Science Foundation. I can’t tell if this is a promotion, demotion, or neither, but I wish him good luck there (yes, he’s read all of your Here’s How To Fix Science Substack posts, so I hope you got them right). 10: Interesting Twitter discussion on the limitations of classical militaries. A kingdom could typically only field one large army, because if the king gave someone else control of an army, they could use it to overthrow the king. The Roman Republic did better - not just because its idea of legitimacy made revolt less likely, but because it had two consuls! 11: Related: one reason for Imperial Roman instability was that low elite fertility prevented the institutionalization of hereditary monarchy. During the period 1 - 250 AD, “only three emperors - Vespasian, Marcus Aurelius, and Septimius Severus - had sons who could inherit.” (h/t @LiorLefineder) 12: Near-Instantly Aborting The Worst Pain Imaginable With Psychedelics - Sasha Putilin gives a good explanation of Qualia Research Institute’s ClusterFree initiative (I’ve signed their open letter; other doctors, researchers, and patients should too). 13: Transformer: The Left Is Missing Out On AI. But not Matt Bruenig, who discusses his opinions here and has released a partly AI-generated book (a collection of details on various important labor law cases - this seems like one of the few examples of where an AI-generated book could be genuinely useful, although I’d be scared of hallucinations). And now he is working on an AI that can answer labor law questions. 14: Also in the important field of Matt-AI interactions: Matt Yglesias: AI Progress Is Giving Me Writers’ Block. “Questions about basically every medium-run policy debate collapse into arguments about the future trajectory of A.I.” - and since we don’t know the future trajectory of AI, it’s hard to write about anything. I have the same problem - remember that the original meaning of “technological singularity” was a point so transformative that it’s pointless to speculate about what happens afterwards. 15: Nicholas Decker: Black Patients Need Black Doctors. Nicholas reviews the literature and finds that it’s not true that racist white doctors mistreat black patients. However, it is true that black patients are less likely to seek help from / cooperate with white doctors, so giving black patients access to black doctors really does improve their health outcomes. A corollary is that exaggerated claims about how white doctors mistreat black patients can really do a lot of harm (by making black patients avoid medical care unless a black doctor is available). 16: Good discussion of ways that looking at benchmarks overestimates Chinese AI capabilities, by @Altimor and @scaling01. 17: In experiments, current AI was not very good at helping a bioterrorism red team with the kinds of tasks involved in creating bioweapons. 18: Every step of this is unsurprising, but the result still wows me - type the name of an object, and an AI will build it for you in Minecraft (click for video): 19: In response to my crime statistics post, Theodidactus discusses the impossibility of truly defining and numbering crimes (eg did Sam Bankman-Fried commit one crime, a handful of different crimes, or [total number of FTX customers] crimes?) Hopefully you can predict m…