2026-04-17 19:11:48
For more than half my life, now, I've wanted to write a good pantoum.
It's bothered me for a long time that most of the famous pantoums in English are not good.[^1] This is in contrast to, say, the multiple phenomenal villanelles; I'm pretty sure I also read a good sestina once, but its identity is tragically escaping me now.
A pantoum is a type of poem based on the Malaysian pantun and with a very rigid repetitive structure: basically, the second and fourth lines of each stanza become the first and third line of the next stanza, and so on. This is the source of the difficulty: each stanza can only introduce two new lines, and those need to be re-usable in the next stanza, without (ideally) sounding either boring, vague or clanky in their new positions.
Readers who hate LLMs please cover your eyes/nose now:
I turned to Claude. "Write a pantoum" has been my personal LLM benchmark for a while now, and whatever developmental leap Claude went through earlier this year means she is now able to write a surprisingly passable pantoum herself.[^2]
But she also gave me two gifts to help me pantoumime on my own behalf. First, she – largely under her own direction – designed this Pantoum Editor that lets you easily see which lines need to match, and where you currently have mismatches, and realign them. I'm really impressed with this product, you can try it for yourself here: http://www.marbiru.com/pantoum-editor.
She also had the patience to argue with me as I crafted the following folly. The first line has been in my head for a decade now, and the writing is all my own, but Claude provided the thing that neither I nor any other human could give me in the meantime, which is some bundle of conscientiousness/structure/motivation to get the thing done. I'll write about this some other time, but I think in some ways it's one of the most interesting use-cases of Claude these days (however long These Days lasts): an enforcement mechanism for people who struggle to enforce themselves.
Anyhow! I do not claim this is the good pantoum I dreamed of, but it does exist, and I'm glad for that.
Structured Poems
You don't write structured poems
because they're good;
you want to prove,
you can.
Because they're good
for nowt – wish
you can
amuse.
For now. Twish!
You want
a Muse,
yourself.
You want,
you want to prove
yourself.
(You don't write).
[^1]: basically: the easy way to write a pantoum is to just make each line vague and mysterious and not entirely grammatical to begin with, then interlock them. You know the kind of thing: "the sky hangs heavy like a labcoat / oh, the supple summon of the waves / my heart beats merry merry every morning / the time of our beginnings is awake." Then the next stanza can start "Oh, the supple summon of the waves:", and then more of the same, and it doesn't make any more or less sense than it did before. This (I'm sorry) is basically what I think of John Ashberry's Pantoum, unless there's something there I'm missing.
Another trick for easier pantouming is to make the subject be something that justifies the repetition. E.g. I think Baby's Pantoum by Anne Waldman is quite good, but it leans heavily on the conceit.
Many English pantoumists have simply agreed you're allowed to bend the words of the lines so long as they're roughly the same. e.g. Parents Pantoum by Carolyn Kizer. I don't understand this, if you relax the constraint you've lost the challenge? (Baby's pantoum cheats a bit as well, I think, but not as egregiously. Also, I am much more sympathetic to cheating slightly after fully internalising how hard it is to do properly).
A big shout out to Nocturne by Sadakichi Hartmann, which I find very pleasing, both because the repeated lines somehow feel altered by their new positions, and because the theme (boats on the water) fits well with the waves of repetition to begin with. Imho this is the best existing pantoum.
[^2]: Tide Work, by Claude
The harbor keeps its ledger in salt,
counting what the moon has loaned the shore.
Boats come back lighter than they left.
Nothing stays where you put it down.
Counting what the moon has loaned the shore,
the gulls audit each retreating wave.
Nothing stays where you put it down—
even the rocks are smaller by morning.
The gulls audit each retreating wave,
patient as bookkeepers in gray suits.
Even the rocks are smaller by morning,
worn to the shape of someone leaving.
Patient as bookkeepers in gray suits,
boats come back lighter than they left,
worn to the shape of someone leaving.
The harbor keeps its ledger in salt.
2026-04-16 19:11:11
Suppose you find out that someone you know did something horrific to someone else. The victim (understandably) doesn't want to go public. What should you do?
It's surprisingly hard for me to answer this question. Morally, the perpetrator should face consequences for their actions, but there's no obvious way to make that happen without also causing consequences for the victim. (I guess you could question the premise that the victim should be allowed to choose whether to go public, but I think the common agreement is that it's their right to choose).
To some limited extent, you can can tell individual other people about the situation, but (depending on context) there's a chance that this still risks outing the victim, and regardless you could only plausibly reach a tiny fraction of the people who need to know.
You can maybe hint vaguely about a person's untrustworthiness in public, but without a concrete reason this is unlikely to be persuasive in general.
I see a lot of stories in the news about people who have done terrible things finally being outed, and it's clear that some non-trivial set of people knew about the behaviour for a long time previously, and then the broader public (understandably) feels extremely upset that these people knew and yet did nothing.
To be explicitly clear, some of the people-who-knew also didn't want to do anything about it, even if they could. Which is its own kind of awful.
But as best I can tell, in many contexts, even if you badly want to do something about it, you can only really do anything once at least one victim is willing to go public, and until then.... you can only look on in horror? (Please tell me I'm missing something, I would badly love to be missing something).
2026-04-15 19:01:22
I think implicitly there are three models of love:
Model 1: you can be equally happy with a very wide range of people – you just need to spend time and grow together.
I think this is the implicit model behind arranged marriages: as long as you pick someone within the (large) pool of people you're basically compatible with it doesn't really matter who it is, you should just settle down with the first reasonable match you get and then do the work of growing to love each other.
(This surely varies by culture, but my understanding is that most modern arranged marriage cultures don't just say "you have to marry THIS person", you get offered matches and you're allowed a couple of "nos" before you say yes. Except that you know that if you say too many "nos" you'll be branded as difficult, so saying "no" is risky. But still, I think this matches the model described above: you're expected to act as if 50%+ of people in your pool are reasonable matches, but you don't have to act like 100% of the pool is good enough).
Model 2: there's a normal-ish distribution of how happy you'll be with different people: some people you'd be very unhappy with, many people you'd be vaguely-ok with, some people you'd be very happy with.
With a normal distribution the tails are very light. Imagine lining up 100 people in a row from the person least-compatible to most-compatible with you. The amount of benefit you get from going from the 50th-best person to the 20th-best is the same as the benefit of going from the 20th-best to the 5th-best, and the same as going from the 5th-best to the 2nd-best (roughly).
But the difficulty of getting that improvement increases each time: it's 2.5x harder to find the 20th best than the 50th best, and 4x harder to find the 5th best than the 20th best, and 8x harder to find the 2nd best than the 5th best.
Unlike the first model, it doesn't make sense to marry the first random person you meet: you really will be happier with your 5th-best person than your 20th-best. But if you're already dating your 5th-best, is it really worth holding out to find your 2nd-best? It depends on your values but I would argue "probably not", given the search costs and the risk of ending up alone.
I think this is, ultimately, the model that most modern WEIRDos have about dating, but would be interested to hear if you think otherwise.
Model 3: There's an extremely fat-tailed distribution of how happy you'll be with different people: the best person for you is meaningfully better than even the second-best person, let alone the 5th-best.
This is the "soulmate" or "one true love" model of relationships. Depending on the numbers, it probably does imply that even if you're currently dating the 2nd-best person for you, it can be worth the trouble to break up and keep searching for the 1st-best, because the happiness that awaits you is beyond comprehension.
The three models have very meaningful implications for how you should live your dating life: from "marry the first person you meet within the top half-ish of your distribution" to "probably hold out for someone 90th percentile or above, but don't fret about getting 99th" to "it is actually very important whether you marry your 99th percentile or 99.9999th percentile person, good luck!"
2026-04-14 19:11:25
James here, the periodic contributor to Atoms vs. Bits. Over the past couple of months, Uri and I have been working with our friend Dr. Dua Hassan on her podcast "Second Opinion with Dr. Dua Hassan".
Dua is a physician at Boston Children's hospital and in the show she takes the time to explain the "Why" behind different medical interventions that a baby or child goes through in a way that you never really have time for in the normal course of business.
The first season of the show is about the first 24 hours of life and the key interventions that most babies have: the Vitamin K shot, the Eye Goop, the PKU test, the Hep B shot, and the first feeding.
I've been through this first 24 hours three times as a father and I like to think of myself as prepared (I read the books before the first baby was born) and I learned a tremendous amount doing the show with Dua, both about the science and about how we decided what the standard protocol for babies should be. The real world is detailed and messy in delightful ways.
Worth checking out if you are a new parent, a soon-to-be parent, or just interested in learning about this corner of the world. You can find it on Spotify here, Apple Podcasts here, or on YouTube here.
2026-04-13 19:11:24
Every so often I'll meet someone who says how the country's problems could be solved if everyone had to do national service at age 18.
My first note is that it (unscientifically) feels to me like these people have never themselves done national service.
My second note is that they're always saying that young people should do national service, and they're always past the age where you'd have to do it.
So here's my proposal: mandatory two year national service, but you do it at age 50.
This would have a large number of benefits versus national service at 18:
1) many people get stuck on their career tracks, and want a "reset" but can't quite get one. Many people say "oh what I'd love to do now is just go become a teacher", but life goes on and you still need to show up at your current non-teacher job and there's never a good moment to retrain and start over. National service for 50 year olds would give an obvious moment for people to switch to a second career.
2) national service for 50 year olds would help disrupt some of our current gerontocracy issues: at a lot of organizations, the younger people never get a chance at management, and (hopefully) this would give some young talent the opportunity to step into bigger roles for a while and prove their capabilities. (I admit that in practice the 50 year olds would just get replaced temporarily by 80 year olds, then come back and reclaim their existing sinecures, but a guy can dream).
3) national service for 50 year olds would allow people to give back to their communities after they've actually developed skills and abilities: imagine how much more service you can serve if you've got 30 years of training and experience under your belt, versus an 18 year old with (at most) enthusiasm.
4) national service is many things, but one of those things is a near-100% tax on selected individuals: while you're doing national service you get paid little-to-nothing, and it's very hard to work elsewhere for money at the same, so it's equivalent to being taxed for almost-all your income for the duration.
Young people generally have less money (on average), and for low-income young people national service can be a really significant burden, delaying their ability to get on a career ladder and earn a reasonable income by multiple years. This is a component that the wealthy older people who (in my experience) advocate national service never seem to acknowledge, and frankly one reason why their blaséness about the whole proposal irks me. But if you believe in progressive taxation, and you're calling for a near-100% tax on anyone, it should obviously be richer people rather than poorer ones paying it.
2026-04-10 19:11:44
Many thanks to all of you who kindly responded to my survey earlier this week about communities. Here were the questions:
(An early version of the survey failed to include "no communities", sorry for those who responded early).
The motivation for this survey was a thought that's chased me around for years, namely: In Imagined Communities, Benedict Anderson claims that newspapers created national identity.[^1]
Specifically, the development of print capitalism meant that suddenly people across the country were reading the same words over their tea and coffee in the morning, and (per Anderson) this massive shared ritual created a sense of kinship across millions of people who would never actually meet each other, therefore creating the modern sense of a nation: an imagined community, because most of the members are strangers, but no less real for being constructed.[^2]
If you've existed in the modern discourse sphere over the last 20 years, you've possibly already had the same thought I did: if newspapers are dying, and (supposedly) we each exist in our own social media echo chamber, do we still have that imagined community? Does Anderson's model provide a framework for the fracture of our polities: we stopped reading the same newspapers over breakfast, so we stopped feeling kinship with our fellow citizens?
Of course, you may also have had the thought "wait weren't there multiple competing newspapers, often aligned with different ideologies" and also "when Anderson talks about this imagined community, was he actually just talking about property owning men of the time, really?", two valid questions which Anderson may or may not answer in the book, it's been a really long time since I read it and I don't remember, sorry.
Also, thinking about it, it's not obvious if the implication of going from "each country reads its own newspapers" to "everyone in the world is using five giant websites filled with screenshots of each other" would lead to more fragmentation, rather than less.[^3] Couldn't Twitter in theory bring everyone together, since we're all reading the same platform every day?
Anyhow. I don't actually have a thesis, but I can at least bring you the results of my survey; here's the types of communities ATVBT readers [^4] feel meaningfully a part of:
My main note is that "political/ideological community" has the most votes. Which doesn't really surprise me, because in my circles I would guess some form of political alignment is the strongly-held identity I most commonly encounter.
Here are some selected quotes of what you all had to say about your communities. I made the survey anonymous so everyone could feel at ease, but if you want to follow up by email please do, I would love to hear more about so many of these responses. I don't have any theories from them, but if you do please do throw down in the comments....
People who care about internal experience in a similar way to how I do
My "other" includes two others:
1. Community of visual artists in my local community and also stretching back through history.
2. Community of people with neurodivergence and learning disabilities.
alas, it is effective altruism
Odd that class is missing. Is it too gauche to admit it? Yes, but this is anonymous so: I (unironically) identify with rootless semi-neoliberal globalists.
Separately, I also feel kinship with people trained in quantitative methods (e.g. economists, engineers, physicists). Quants vs non-quants is an under-appreciated cultural divide. I am not a STEMlord*; it just functions like the options above in that we see the world in similar ways. * I am especially not a “rationalist.” They’re mostly quant cosplayers.
I feel kinship toward the Jewish community even though I don’t practice the religion as a result. It’s more a cultural kinship than religious.
I feel a kinship with people from my home state (CT), I think mostly because I don’t live anywhere near it anymore so it’s rare to meet someone else from there. Almost like being an expat.
I also feel kinship toward supporters of some sports teams I support (go Mets!), possibly for similar expat-esque reasons.
It seems to be when I’m in a minority, I feel more kinship with people in that minority with me.
A sober community (alcohol free, other than AA)
Community of literature lovers
Brown Women
Hate to say it but financial or work peer group is also my community because we have the same or similar lived experiences in the past 10-20 years
Well, to be honest, none of the above or any other not in this list! And it’s taken a ton of careful effort to get to this point.
[^1]: Assume all my claims about what Anderson does or doesn't say are kinda rough, it's been 15 years since I read the book. I tried to re-read it in order to write this post, actually about a year ago, and kept failing to do it for so long that I've given up and just decided to write this without re-reading him, sorry.
[^2]: not to get political on you, but one of his claims (as I recall it) is that many national conflicts involve people saying "the other side's national identity is imaginary, and therefore not real!", and Anderson's points are a) just because it's imagined doesn't make it fake, and b) your side's national identity is also imagined in the same way, you just don't realize it. [Of course, I'm sure he wasn't talking about YOUR national conflict!]
[^3]: I do greatly enjoy the "five giant websites filled with screenshots of each other" quote (h/t someone called Tom Eastman, apparently), and think of it often, but I also want to point out that it's not true: this is just the Western/American-sphere internet, and there are at least 2 other parallel ecosystems in the Russosphere and the Sinosphere where they use a different set of websites.
[^4]: or at least, those who filled out the survey, which was about 2/3 of those who clicked the link. Goes without saying that "readers of this blog" is not a representative sample of anything, but, you know what they say, you go to survey with the sample you have.