sarahconstantin

Karma: 9,446

sarahconstantin 12 May 2026 14:11 UTC
3 points
0
on: sarahconstantin’s Shortform
links 5/12/26: https://roamresearch.com/#/app/srcpublic/page/05-12-2026
- - https://www.worksinprogress.news/p/pancreatic-cancer-just-met-its-match Ruxandra Tesloianu on daraxonrasib, the new drug that Actually Works on pancreatic cancer.
  - it targets RAS, which is an old familiar oncogene, but what i hadn’t realized is it’s “undruggable” in the traditional sense; no pockets!
    cryo-EM mapping of the protein let them come up with a “molecular glue” drug that attaches RAS to another molecule, cyclophilin A, rendering it inactive. i always wondered if structural biology actually had medical applicability, but here’s an example of it being essential and useful—we wouldn’t have had this drug but for a precise molecular map of the target molecule!
  - https://en.wikipedia.org/wiki/Kirpan Sikhs are religiously obliged to carry a dagger; “The tenth and final guru, Guru Gobind Singh formally included the kirpan as a mandatory article of faith for all baptised Sikhs, making it a duty for Sikhs to be able to defend the needy, suppressed ones, to defend righteousness and the freedom of expression.”
  - https://altdotbinaries.com/2026/05/11/metis-vs-techne/ this is a thoughtful essay about James C. Scott, but I am an exception to the rule, apparently, because I like High Modernism and legibility; i see myself in the position of someone who would like to analyze data or someone who will be out of the loop unless it is made legible to me, more often than I see myself in the position of someone trying to escape the Powers that Be by being confusing.
  - the other side of “if i am legible, i will be oversimplified, flattened, and controlled” is “if you are illegible, I cannot make head or tail of you; I am forever an outsider, a novice, a non-initiate; you have made your world inaccessible, except to your cronies who know the secret handshake; I cannot help you, I cannot win your trust; I cannot learn anything, I cannot get anything done; you have stalled my progress, mired me in endless inconvenience, presented unfriendly barriers; you have blocked and thwarted me out of spite, even at cost to yourself; and you say it is because you are “truly human” and I am not! you say I am ignorant because I want to know, that I can never understand you because I want to understand you, that I am a machine because I use machines! I can’t even express the perversion that is not wanting to share information, not wanting to get things done, not wanting transparency and efficiency; it’s backwards, it’s the opposite of being an agent, it’s something any child would naturally chafe against, it’s the dead hand of stasis and corruption, it’s letting fear get the best of you, it’s despair of cooperation being possible, it’s against goals and wants and will itself...”

sarahconstantin 11 May 2026 17:25 UTC
3 points
0
on: sarahconstantin’s Shortform
links 5/11/26: https://roamresearch.com/#/app/srcpublic/page/05-11-2026
- - https://en.wikipedia.org/wiki/Wabenzi African kleptocrats
  - https://www.quarter—mile.com/The-Locals-Dont-Know fair enough; living in a place is not at all like being a tourist. locals have a daily routine. they probably aren’t exploring all the most fun things to do. when tourists come to my city, i’m usually embarrassed about how little I can tell them about it.
  - https://defensesindepth.bio/10-big-projects-for-reducing-bio-x-risk/ these seem fine, and I applaud making a “wish list” in biosecurity public. i’d have expected a lot more...technical meat, though? like, nitty-gritty with specific threat models of how engineered pandemics are expected to work and what techniques will counteract them. these ideas seem to be based on a belief that we’re very close to doomed when it comes to technical or policy safeguards, or even most types of preparedness. if you’re talking about DIY respirators, your assumptions are extremely pessimistic—maybe too pessimistic! yes, that sort of thing was useful in COVID, but COVID was a complete snafu. do you think we don’t have time to stand up a more functional institution before the Next Big One hits???
  - https://en.wikipedia.org/wiki/Redeemers perhaps not everyone knows this story
  - https://en.wikipedia.org/wiki/Annie_Oakley her life was much more interesting than i realized. orphaned. horrific childhood. learned to shoot so she wouldn’t starve. cutest marriage ever (he fell in love when she beat him at a marksmanship contest). taught 15,000 women to shoot, firmly believed that firearms were important for women’s independence.
  - https://en.wikipedia.org/wiki/Calamity_Jane by contrast...she seems like kind of a loser? it’s not clear why she’s famous at all, apart from featuring in some highly fictionalized Westerns.
  - https://en.wikipedia.org/wiki/Thomas_Berger somehow i forgot all about him. interesting era in literature.
  - https://everything2.com/node/e2node/Freedom%20of%20the%20press%20and%20repression%20of%20the%20photocopier#google_vignette in Sweden, “freedom of the press” refers very literally to things printed on printing presses; there are fewer rights granted to things that are photocopied, handwritten, posted online, etc.
  - https://www.lesswrong.com/posts/cWFsCFyCttsiJwn2j/bringing-more-expertise-to-bear-on-alignment this is about connecting AI alignment to more “straight machine learning” research fields such as complexity theory, statistical learning theory, “computational mechanics” (the geometry of representations in neural networks), “singular learning theory” (mapping properties of a model to properties of its training data), and more. I think this is promising, if-and-only-if we aren’t trying to prevent a catastrophe in like <2 years.
  - as an “AI moderate”/”foom skeptic”/whatever, i think it’s entirely possible that it makes sense to start multi-year, heavily theoretical alignment projects, in addition to less-theoretical, more-urgent stuff around near-term misuse or accidental-catastrophe risks (policy, evals/monitoring, resilience stuff like “level up your cybersecurity” and “try to make formal verification real”).
    however, if you think we are literally going to die before 2030, starting a theoretical research program in a new paradigm is clearly futile.
    my intuition is also that it’s...sort of salutary? and neglected? to try to do more “serious” and ambitious theoretical research. like, people would be tempted to not bother with that even if it were a good idea. that’s not in itself a reason to do theory, but it’s on my mind.
  - https://arbresearch.com/files/Embedme.pdf man i wish this were on the web so i could check it out
  - https://arxiv.org/pdf/2508.14390v1 this is a bad paper (uses out of date models, narratizes results not in the data) but it does engage with the question of whether AI models are calibrated in the sense that they’ll “bet harder” on questions they are in fact more likely to get right, if instructed to maximize their expected value in a betting game. (this is somewhat different from asking them to give their confidence as a percentage, from a prompting/persona perspective.) the answer is that Llama3.1 and Qwen2.5 are not calibrated on MMLU-pro, but are on the easier Trivia-QA...
  - https://arxiv.org/pdf/2311.08298 mostly eliciting confidence/calibration from AIs focuses on how to get them to be more calibrated, or whether you can use white-box methods like training a model or using logits to predict their odds of getting things right. I’m interested in calibration as an onramp to thinking about machine consciousness (“do they know what they know?”) but the field isn’t so much asking questions that would be useful for that.
  - https://arxiv.org/abs/2207.05221 “Language Models (Mostly) Know What They Know” is an awfully confident title for a paper that does not fucking prove that.
  - https://arxiv.org/abs/2601.17406 analyzes characteristics of AI-generated code from different models. mostly this is about language in the PR rather than in the code itself; one exception is that Claude Code has more conditional statements per line of code than other agents.
  - https://arxiv.org/abs/2602.12413 AI benchmarks overstate AI progress because of “soft contamination.”
  - https://commoncrawl.org/ important public resource!
  - https://news.ycombinator.com/item?id=48055261 measured in rice, as the feudal Japanese did, anyone making over $300k/year is as rich as a daimyo, one step down from the shogun.
  - https://blog.cloudflare.com/building-for-the-future/ layoffs in critical internet infrastructure are Concerning, from the point of view of preventing outages etc.
  - i’m inclined to believe that large rounds of big-company layoffs do not, in real life, happen when it’s economically rational, but that there are times when it’s socially unacceptable to do em, and times when it’s socially acceptable. so you get these big waves when everybody’s doing layoffs. it wouldn’t be so correlated if companies were continuously optimizing the correct profit-maximizing # of employees. and right now “AI” is an excuse to do layoffs. it may also be rational to reduce headcount because of AI, in some companies, but I think the actual considerations are more complicated in practice. and there’s a lot of spray-and-pray in practice, where layoffs are random or correlated with something irrelevant to performance, instead of more-productive employees getting preferentially retained.
  - https://www.lawfaremedia.org/article/kicking-the-tires—a-voluntary-path-to-pre-deployment-ai-vetting this is Dean Ball’s policy agenda on AI

sarahconstantin 7 May 2026 16:22 UTC
3 points
0
on: sarahconstantin’s Shortform
links 5/7/26: https://roamresearch.com/#/app/srcpublic/page/05-07-2026
- - https://meaningness.substack.com/p/creating-conduits-visionary-actual David Chapman on how yidam practice actually works. Fascinating, and tempting!
  - I do have a sense of caution about what is essentially a polytheistic practice; I’m an atheist who’s generally fine with engaging in woo in a spirit of experiment, but I have a vague sense that dealing with gods is probably frowned upon by monotheists for good reason. It might do something to you that you wouldn’t want. In the same way that some people argue IFS can “give” people something more like multiple personalities when it’s healthier to be unified, I wonder if dealing with multiple gods might “break up” your sense that you live in one world/one reality. (A friend of mine used to joke that atheists believe in one god, Athe.)
  - https://frontier2025.netlify.app/ recent scientific progress
  - https://arxiv.org/abs/2308.10248 steering vectors in AI were an Arb Research project
  - https://openresearchinstitute.org/onboarding/A_B_U.html I like this A/B/U system.
  - When I encounter a claim, I can’t usually easily sort it into “I believe this” or “I don’t believe this”. There are other buckets, like “all I know is this guy said the thing, I cannot or have not compared it to my other beliefs to see what I think of it”. Epistemic systems need an “other” bucket, not “I think it’s true”, “I think it’s false” or even “I am uncertain”, but something like “unprocessed”. Most claims I read are unprocessed! (Certainly most upsetting claims go in the “gee, that would suck, wouldn’t it? i can’t stand to think about it!” bucket.)
    I also like that U stands for “unclear, undefined, unknowable, or untrue”—i.e. anything where I don’t clearly recognize the truth of it. it’s caused a lot of anguish for me over the years when I can’t rule out that a claim might be true. In A/B/U, you don’t actually have to distinguish “true” from “false” every time; you distinguish “clearly true” from “everything that’s not clearly true”. Thinkers who throw up a lot of Us and not a lot of As or Bs (novel or familiar truths) are not necessarily wrong, but you’re not in a position where you should rationally trust them very far. (And, to put this more reassuringly, you are free to not trust them very far, you are not epistemically obliged to treat them as correct or to keep investigating until you can “disprove” their theories.)
  - https://en.wikipedia.org/wiki/Michel_de_Montaigne Fascinating guy. Marrano background! Humanist childhood!
  - https://en.wikipedia.org/wiki/Battle_of_the_Teutoburg_Forest this is what “Varus give me back my legions” is from.
  - https://conclave1492.com/ want.
  - https://www.reddit.com/r/Welding/comments/q57fpq/welders_of_reddit_is_anyone_in_a_shop_that/ apparently working with a welding robot is mostly about “feeding” the robot pieces
  - https://academic.oup.com/jeea/article-abstract/19/6/3104/6179884 as of 2021, robots do not reduce manufacturing employment in Germany; they shift it towards programming roles.
  - https://www.aiprm.com/robotics-statistics/ half of robots are used in “handling”. (such as pick & place.)
  - https://www.engineering.com/the-what-why-and-how-of-roboforming/ roboforming deforms a sheet of metal with a robot arm and steel ball; it’s a tooling-free replacement for stamping.
  - https://www2.census.gov/library/working-papers/2023/adrm/ces/CES-WP-23-14.pdf where robots are adopted
  - https://blog.atlascomputing.org/p/civilizations-maintenance-backlog good ideas for AI resilience
  - https://www.essentialtechnology.blog/cp/178994713 “Securing Critical Components of Cyberphysical Systems.” Seems like pure common sense with regards to software (I can’t evaluate the hardware claims.)
  - https://claude.ai/share/b675fa33-35a1-4aa7-a044-ea877ff4e643 discussion with Claude about supply chain risks from a world where AI agents do the purchasing. the “bullwhip effect” may be amplified, causing shortages and gluts.
  - https://arxiv.org/html/2502.11620v3 you can upper-bound AI-generated code correctness by looking at how much program behavior varies with rerolls of the same prompt. you can map programs into semantically equivalent clusters. if an AI agent produces lots of different clusters, they can’t all be correct, so any one of them is less likely to be correct.
  - https://en.wikipedia.org/wiki/Guido_of_Siena good example of 13thc Sienese painting
  - https://messymatters.com/autonomy/ how Bethany Soule and Daniel Reeves stay financially independent while married. I’d never do this but it’s cool.
  - https://modernpower.substack.com/p/taking-stock-of-abundance-may-2026 state of the Abundance movement; much to consider.

sarahconstantin 4 May 2026 16:34 UTC
4 points
0
on: sarahconstantin’s Shortform
links 05/04/26: https://roamresearch.com/#/app/srcpublic/page/05-04-2026
- - https://www.hyperdimensional.co/p/aviate-navigate-communicate Dean Ball has good ideas about governance for Mythos-style models that create security vulnerabilities.
  - yes, if we’re going to regulate AI capabilities, the government should pick a rule and stick to it, not permit models from its friends and forbid models from its enemies in the ad-hoc fashion the Trump Admin has been. Also, it probably makes more sense to have independent, government-licensed private organizations do the model evaluations than to do it within tiny CAISI.
  - https://jobs.ashbyhq.com/Merge%20Labs/29ff7e5f-0621-48c2-91d1-e9bfe60f505c “neuron whisperer” is a great job title for someone who specializes in neuronal culture. cells love to die, and no cells love to die as much as neurons. i hope they find their neuron whisperer.
  - https://reason.com/volokh/2026/05/03/how-european-libertarians-differ-from-american-ones/ European libertarians/classical-liberals seem neat.
  - https://en.wikipedia.org/wiki/Synoptic_Gospels a catchall term for Matthew, Mark, and Luke, the Gospels that mostly cover overlapping narratives of Jesus’s life.
  - https://jonahsinick.github.io/faculty-faces/ranked.html the faces of different academic fields
  - https://www.bitsaboutmoney.com/archive/nonprofit-indicted-bank-fraud/ damning stuff about the SPLC and associated nonprofits, by Patrick McKenzie at his most acid. yes, it looks like they plainly broke the law by lying to banks about the shell companies used to pay hate group informants. Also, they may well have broken the law in additional ways, by campaigning to get fund transfers to Trump-supporting political donation groups blacklisted. (nonprofits are not allowed to take sides on political campaigns.)
  - In general, I’m sympathetic to private boycotts or corporate policies against speech they consider bigoted or otherwise inappropriate. Free speech and free association should protect private individuals and organizations who object to certain views and choose not to do business with those who express them. And there are views I find objectionable and prefer not to encounter, with different standards for different contexts.
    but the campaigning behavior being described here seems legitimately horrible. this was not an organic expression of public distaste. it was also not a set of independent choices by corporate executives based on their own values. it was a manipulative and gross organized campaign. even where this sort of behavior is legal, it violates some of my strongly held intuitions about “minding one’s own business”. it is really not some activist group’s business to tell corporations who they may and may not platform or allow to access financial services! you can have an opinion, but you can’t presume to give orders when it’s not your company and not your money. and no business should be taking orders from people who, properly, are just bystanders who happen to have opinions; it’s astonishing that they did. Something very screwy went on here, and I think I understand some of the accusations over the past several years better than I used to.

sarahconstantin 20 Apr 2026 16:17 UTC
6 points
0
on: sarahconstantin’s Shortform
links 4/20/26: https://roamresearch.com/#/app/srcpublic/page/04-20-2026
- - https://www.lieder.net/lieder/get_text.html?TextId=91385 this is the song in Hilary Mantel’s Thomas Cromwell trilogy
  - https://en.wikipedia.org/wiki/Nicholas_Culpeper the herbalist was related to the earlier courtier Thomas Culpeper
  - https://thezvi.substack.com/p/on-dwarkesh-patels-podcast-with-nvidia satisfying rundown, I agree with Zvi Mowshowitz on all points, as i often do
  - https://www.lesswrong.com/posts/WewsByywWNhX9rtwi/current-ais-seem-pretty-misaligned-to-me this seems weak to me. the kind of “misalignment” that is “AI coding agents are too eager to say they’ve solved your problem when they really haven’t” is a.) exactly the sort of thing AI labs are economically incentivized to fix, and b.) kind of unreasonable to complain about?! if you are letting your agents completely off the leash on open-ended projects, they are going to screw things up, and you will need some kind of supervisory process. it seems to me like serious professionals already know that and are successfully compensating for it? if you think of the models as TOOLS YOU USE, which of course have flaws and limits, but are trending in the right direction over the course of months and years, you tend to have a better time with them and make predictions more in line with reality...
  - https://www.verysane.ai/p/alignment-is-proven-to-be-tractable i’m not sure i agree with the *title* but i agree with the essay, i was surprised by how little of a thing “brittleness” is in large 2020s models (in the sense of “AI model does something very different from what the user intended” or “model doesn’t generalize to slightly different real-world scenarios from the training data”). language seems to really help with that!
  - https://en.wikipedia.org/wiki/List_of_human_cell_types red blood cells are by far the most common cell type in the human body

sarahconstantin 8 Apr 2026 15:53 UTC
2 points
0
on: sarahconstantin’s Shortform
links 4/8/26: https://roamresearch.com/#/app/srcpublic/page/04-08-2026
- - https://benjaminrosshoffman.com/telescopes-need-good-lenses/ this amounts to “yes, I think I am a better parent than liberals/progressives.” I am unconvinced.
  - One of his examples is that his kids are in the 99th percentile for height and weight despite not being visibly fat, thus he thinks that other parents are underfeeding their children. This doesn’t make sense. Child obesity is in fact a significant problem; on average, US parents probably feed their kids too many calories, not too few.
    The fact is that—I don’t know why, maybe measurement error at the doctor’s office, which is common—it’s not that unusual for kids to be “outliers” in size. Anecdotally it seems to me like “99th percentile” large or “1st percentile” small kids (who appear neither fat nor scrawny) are a lot more than 2% of the population. the thing is, the official distribution of heights and weights for kids is very narrow, so a “99th percentile” kid is not that enormous, and a 1st percentile kid is not that tiny. Plus, kids’ growth rates change enough over time that they usually grow out of “outlier” status. IMO, Ben’s kids are normal-sized, their “99th percentile” status may very well be random fluctuation unrelated to how he feeds them, and he is really jumping to conclusions from inadequate data to conclude that most of his peers underfeed their kids.
    I’ll grant him that being better at cooking tasty food might make kids less likely to be picky eaters, but there’s a lot of noise in that as well (I’m a pretty good cook and my kids often don’t like what grownups think is delicious). And I also am skeptical that liberals are worse cooks than conservatives on average in the US. (remember, the OP Scott post was about politics!)
    I have less objection to his overall point, which is a critique of altruism itself. Yes, people will never be as careful or as grounded, as effective, in helping others, if they’re motivated primarily by getting good-boy/good-girl points, as if their practical self-interest is entangled.
    I’m not sure that I buy that EA-style or progressive-style altruism is outright bad; for the most part, i think people aren’t harming themselves by doing it, and even if they help suboptimally they may still be helping a little. I see people as mostly putting themselves first, with a little ineffectual flavor text to make them feel good about themselves, and maybe also a little genuine help going to people who need it, and I don’t see a problem with the flavor text or the modest amount of help. I think charity is not obligatory and not the most important part of morality, but it’s probably net good; and I’m not too fussy about whether people rationalize/exaggerate how good they are, if it makes them feel better. (I also think building AI is cool and have no moral objection to it, so I don’t hold EAs blameworthy there either.)
    I don’t like the idea that we evaluate the goodness of “communities” by how “functional” they are, including via measuring fertility—yes, in an evolutionary sense, fertility is “fitness”, but that doesn’t tell me why i should care about it. I don’t want to be Satmar or Amish. If they are “virtuous”, so much the worse for virtue!
  - https://icmi-proceedings.com/ICMI-E-virtue-under-pressure.html Tim Hwang is trying to make AIs with Christian values. this freaks me out, even though i suppose it shouldn’t (Christians are not so overwhelmingly dominant worldwide that we wouldn’t also have secular AIs)
  - https://www.newyorker.com/magazine/2026/04/13/sam-altman-may-control-our-future-can-he-be-trusted entertaining and thorough profile of Sam Altman, including a very well documented pattern of people saying he lies a lot and can’t be trusted, going back to his YCombinator days.
  - https://nymag.com/intelligencer/article/why-citrini-research-sent-an-analyst-to-the-strait-of-hormuz.html I gotta admire a finance firm that sends a researcher to an active war zone to get firsthand data.

sarahconstantin 6 Apr 2026 16:36 UTC
2 points
0
on: sarahconstantin’s Shortform
links 4/6/26: https://roamresearch.com/#/app/srcpublic/page/04-06-2026
- https://balioc.tumblr.com/post/768319203871719424/looking-at-the-world-from-a-managers-perspective I like this taxonomy. I’m definitely a Fanatic.
- people’s vulnerability to sleep deprivation is a bona fide Trait: there’s substantial variability across people, substantial consistency within an individual’s sleep-dep-vulnerability over time and across measurement types, and substantial heritability. some people are more messed up by sleep deprivation than others!
- - https://pmc.ncbi.nlm.nih.gov/articles/PMC3413799/
  - https://academic.oup.com/sleep/article/43/6/zsz292/5648124
- https://german.millermanschool.com/ fun quiz

sarahconstantin 2 Apr 2026 17:51 UTC
3 points
0
on: sarahconstantin’s Shortform
links 4/2/26: https://roamresearch.com/#/app/srcpublic/page/04-02-2026
- https://www.astralcodexten.com/p/against-the-concept-of-telescopic
- - I basically agree based on personal experience that telescopic altruism is not a thing. Liberals, progressives, and EAs who care about distant strangers don’t especially neglect their own family and friends. It’s just not a pattern I observe at all.
  - It seems like everyone puts themselves and their family/friends first. and to the extent “lefties” even do tend to care more than “righties” about distant strangers, that care doesn’t seem to come out of a fixed budget such that “righties” do nicer things for their family & friends. the “extra caring about strangers” is either virtually free (it affects how you vote, or how you talk, or what you read, or what you work on, and “righties” also vote/talk/read/work, just differently) or it just makes you generically more worried, because you worry about distant people in addition to nearby people.
  - I do think there is a thing where some people put energy into following politics at the expense of their career or their housework, which is a sort of “caring about distant strangers more than matters close to home”, and can be irresponsible… but that seems to exist on both sides of the political spectrum!
- https://www.myjewishlearning.com/the-nosher/does-the-great-british-baking-show-know-what-challah-is/
- - the Great British Baking Show is laughably wrong about what challah is. are the Brits okay?
- https://medium.com/supernuclear/fairness-is-overrated-and-bragging-is-underrated-motivational-systems-for-coliving-and-beyond-c83acf9f88c9 good ideas for motivating people to do chores in group living situations (including families)

sarahconstantin 31 Mar 2026 17:44 UTC
3 points
0
on: sarahconstantin’s Shortform
links 3/31/26: https://roamresearch.com/#/app/sarahbrain/page/03-31-2026
- - https://writing.antonleicht.me/p/press-play-to-continue a cogent explanation of “why not Pause AI, given that you are concerned about AI risk?” pretty convincing to me.
  - tl;dr: you don’t want to support bad policies that don’t affect x-risk and/or are unlikely to pass
  - https://www.complexsystemspodcast.com/episodes/delve-into-compliance-theatre/ Patrick McKenzie on Delve. he believes they’re probably helping startups fake compliance paperwork. this is illegal and will get a lot of people in trouble.
  - https://en.wikipedia.org/wiki/Rule_against_perpetuities you cannot exert control over private property for more than 20 years after the owner’s death.
  - https://summerlightning.substack.com/p/the-two-memories AI and embodied vs. contemplative memories. ofc i’m mostly on the contemplative side.
  - https://acoup.blog/2026/03/25/miscellanea-the-war-in-iran/ a case that this war is a bad idea
  - https://www.programmablemutter.com/p/gooning-towards-the-fuhrer-as-policy rather satisfying (too satisfying?) theory of how the Trump administration operates
  - https://www.overcomingbias.com/p/finish-the-industrial-revolution Robin Hanson on his usual obsession. i have to ask myself, what would I be willing to sacrifice about modern liberal democratic culture in order to make it out-reproduce the Amish or Haredi in the long run… that wouldn’t make it worse, by my values, than the Amish or the Haredi? like, if you buy the argument that in the long run the future is dominated by Malthusian cultures that sacrifice personal freedom and per capita consumption to reproduce more, and you want to turn your own culture into a Malthusian one, maybe the result is not going to be a nicer culture to live in than other Malthusian cultures. Maybe the problem is precisely the Malthusianism and, if i have to pick my flavor, i might pick Jewish Malthusianism as the least-worst option.

FW26 Color Stats

sarahconstantin14 Mar 2026 15:50 UTC

21 points

0 comments2 min readLW link

(sarahconstantin.substack.com)

Things that Go Boom

sarahconstantin13 Mar 2026 19:00 UTC

61 points

2 comments8 min readLW link

(sarahconstantin.substack.com)

sarahconstantin 10 Mar 2026 16:05 UTC
10 points
0
on: sarahconstantin’s Shortform
links 3/10/26: https://roamresearch.com/#/app/srcpublic/page/03-10-2026
- https://manufacturing-operations.vercel.app/index.html Ben Reinhardt’s matrix of manufacturing operations, ranked by how valuable closed-loop self-improving automation would be
- bottlenecks in defense production:
  - https://www.businessdefense.gov/ibr/mceip/dpai/dpat3/ DPA Article III is a department involved in procuring supplies from the private sector
  - https://www.gao.gov/assets/gao-18-45.pdf solid rocket motors, used in missiles, are in shortage; carbon fiber motor casings are one of the bottlenecks
  - https://www.csis.org/analysis/first-battle-next-war-wargaming-chinese-invasion-taiwan wargames of an invasion of Taiwan
  - https://www.insensitivemunitions.org/history/the-introduction-of-castable-plastic-bonded-explosives-pbxs-in-navy-munitions/ plastic bonded explosives are also a key bottleneck
  - https://en.wikipedia.org/wiki/2025_Tennessee_manufacturing_plant_explosion explosives manufacturing is quite dangerous
  - https://www.navalsteminterns.us/nreip/labs/nswc_indianhead.html missile warheads are made and tested at a single naval facility
  - https://en.wikipedia.org/wiki/Holston_Army_Ammunition_Plant this is the only place where the explosive material for missiles is made
  - https://www.heritage.org/tidalwave/chapters/chapter-6-assessing-the-us-indo-pacific-munitions-system we are short of missiles by about an order of magnitude, Heritage estimates
  - https://www.defensenews.com/space/2017/06/21/us-industrial-base-at-risk-for-key-rocket-motor-ingredient/ there is only one supplier of a key ingredient for rocket oxidizer, ammonium perchlorate
  - https://nuclearcompanion.com/data/mark-48-torpedo/ parts of a torpedo
  - https://www.globalsecurity.org/military/systems/munitions/lrasm.htm the LRASM, a key missile
  - https://news.usni.org/2026/01/24/2026-u-s-national-defense-strategy the US’s official current strategy in the Pacific: “denial defense along the Pacific Island Chain”, i.e. preventing an invasion of Taiwan by being ready to sink any Chinese fleet attempting to attack
- https://dynomight.net/pattern/ Dynomight, food for thought
- https://catboost.ai/docs/en/ library for gradient boosting
- https://paulgraham.com/brandage.html Paul Graham on watches as an extended metaphor for what happens when technical competence is commoditized. he never mentions AI but it’s interesting to consider what the analogies would be.
- https://thezvi.substack.com/p/anthropic-officially-arbitrarily Zvi Mowshowitz, like all reasonable people, decries what’s going on here
- https://timhwang.github.io/because-china/ China is an excellent excuse for whatever you wanted to do anyway
- https://en.wikipedia.org/wiki/Distillation#History the Babylonians had distillation, but for perfume, not booze
- https://en.wikipedia.org/wiki/History_of_gunpowder it is entirely a myth that the Chinese used gunpowder only for fireworks and were unaware of its military applications. they used gunpowder weapons for centuries before Europeans did.
- https://en.wikipedia.org/wiki/Control_variates method used in Monte Carlo estimation

sarahconstantin 2 Mar 2026 20:30 UTC
8 points
0
on: sarahconstantin’s Shortform
links 3/2/26: https://roamresearch.com/#/app/srcpublic/page/03-02-2026
- https://arxiv.org/abs/2602.07164 more effective way to elicit “in character” personalities from LLMs
- https://blog.egan.bio/p/reviving-estrogen-replacement-therapy overview on why hormone therapy for menopause is losing its black-box label (tl;dr: it’s not a totally unreasonable choice)
- https://www.owlposting.com/p/neurotechnology-for-cancer-ben-woodington Abhishaike Mahajan interview with the founders of Coherence Neuro, which uses neuromodulation to disrupt (mostly brain) tumors. neurotransmitters can be tumor growth factors!
- https://en.wikipedia.org/wiki/Fixed-income_attribution#Modeling_the_yield_curve functions that define the yield curve for bonds
- https://lydialaurenson.substack.com/p/social-contracts-moral-order-weaving Lydia Laurenson on “missing stair” type dynamics, how hard it is to enforce norms against bad actors in a community
- https://www.hyperdimensional.co/p/clawed Dean Ball is sticking his neck out to do the right thing here
- https://www.clinicaltrialsabundance.blog/p/ai-wont-automatically-accelerate sorry Dario Amodei, AI can’t make all clinical trials take under a year; sometimes you need follow-ups longer than that for the actual course of disease to become clear! this is crazypants and we need to keep in mind the actual places there’s room for improvement on trial times—reducing purely regulatory delays, or procedural time (like patient recruitment time) as opposed to medically-relevant follow-up time, and finding surrogate endpoints where good ones exist. (I kinda hate surrogate endpoints cause so many suck, but yeah, clearly, when you find a good one it will speed up trials substantially.) Ruxandra Tesloianu bringing the common sense and much-needed urgency, as usual.
- https://marginalrevolution.com/marginalrevolution/2026/02/if-you-have-the-right-to-die-you-should-have-the-right-to-try.html right-to-try is the least-good and most popular form of medical deregulation, IMO, but it is still correct. Ruxandra Tesloianu again.
- https://read.technically.dev/p/how-are-companies-using-ai survey of how companies are using AI
- https://www.economicforces.xyz/p/we-dont-need-to-just-make-up-fantasy solid economics based takedown of the Citrini piece
- https://www.gleech.org/enhance I think I buy this, by Gavin Leech; nobody will ever acknowledge that the transhumanists were right, they will simply adopt “better-than-well”/human-augmentation technologies piecemeal, as those technologies become practical, safe-enough, and accessible.

Getting Back To It

sarahconstantin26 Feb 2026 20:30 UTC

38 points

1 comment7 min readLW link

(sarahconstantin.substack.com)

sarahconstantin 20 Feb 2026 16:45 UTC
32 points
0
on: sarahconstantin’s Shortform
- https://danfrank.ca/daniel-isms-50-ideas-for-life-i-repeatedly-share/ Dan Frank’s life advice.
- https://notnottalmud.substack.com/p/the-tasting-day-why-buying-5-babkas Dan Frank advises holding a tasting as a genre of social event. i should do this more.
- https://www.ribbonfarm.com/2022/03/30/fermi-estimates-and-dyson-designs/ Venkatesh Rao’s idea of a “dyson design” is the sort of quantitative/heuristic thinking that involves exponentials, order-of-magnitude estimates, and similar types of “this is so big that we can neglect that” types of reasoning. Moore’s Law (as a description of the world, and implications drawn from it) is in this category.
- https://www.science.org/doi/10.1126/science.aea1260 universal vaccines? mucosal vaccination in mice with some immune-stimulating antigens protects against multiple viruses & bacterial infections
- https://www.slowboring.com/p/why-are-young-liberals-so-depressed
  - one of my pet peeves is that people usually condemn negativity in negative terms, to the effect of “nobody likes it when you’re a whiny downer.” i don’t believe this is an effective way to fight negative affect! it violates basic principles of vibe!
  - Yglesias instead does this the right way, via inspiration. “remember We Shall Overcome, Si Se Puede, and There Is A Power In A Union? isn’t hope and optimism and working for noble causes better than depression?”
- https://marginalrevolution.com/marginalrevolution/2026/02/rebuilding-our-world-with-reference-to-strong-ai.html I’m pretty much with Tyler Cowen on this; AI will be cataclysmic, on a historical scale.
  - now, given that he believes this, i think it would be reasonable of him to outwardly panic a lot more. if things are probably going to be Quite Bad and it’s unlikely anyone can do anything about it, you’re free to have whatever reaction you like, including putting it out of your mind and spending your time on pleasant things, or running around panicking like a chicken with its head cut off. it’s hard for me to see the point of sober Smart Guy talk, under potentially cataclysmic conditions; it’s so much aversive effort for so little reward.
- https://danluu.com/julialang/ the Julia language is buggy
- https://alexaraki.substack.com/p/how-to-feel-invisible known goodguy Alex Araki wants to build a blood test for [[chronic pain]]. Is there a good molecular correlate of pain? idk but i hope so. new pain drugs have just the worst track record, and an objective outcome measure might help with that.
- https://austinvernon.substack.com/p/speed-can-reindustrialize-america remarkable claim from Austin Vernon that we don’t need industrial policy or new automation tech to reindustrialize America—we just need machine shops that use modern office software for things like quoting, billing, and other “soft” functions (not the machining itself.) He argues that American machine shops could scale to be big—they’re not inherently mom & pop operations—with the right operational setup (eg modern logistics). includes a list of companies he thinks deserve investment.
- https://en.wikipedia.org/wiki/Gantt_chart why didn’t Napoleon just use Gantt charts for logistical planning? because they wouldn’t be invented for another hundred years. and they were instrumental in WWI.

sarahconstantin 11 Feb 2026 17:21 UTC
5 points
0
on: sarahconstantin’s Shortform
links 2/11/26: https://roamresearch.com/#/app/srcpublic/page/02-11-2026
- https://www.clinicaltrialsabundance.blog/p/the-moderna-rtf-and-the-cost-of-regulatory Ruxandra Tesloianu on why FDA changing its mind to reject trial protocols it previously approved is harmful for biomedical progress
- https://www.drmaciver.com/2015/04/some-empirically-derived-testing-principles/ D.R. Maciver advice on software testing
- https://notebook.ldeming.com/garden/index.html Laura Deming on how, if longevity ever really works, it would be creating a new sort of creature with a mindset very different from a “normal”-lifespan human
- https://mcfunley.com/choose-boring-technology advice on picking a software stack
- https://www.owlposting.com/p/heuristics-for-lab-robotics-and-where excellent overview of new directions in (biology) lab automation by Abhishaike Mahajan
- https://authorshipai.vercel.app/ writing app that’s more of a thinking partner than just writing for you
  - I don’t care for AI assistance in writing myself, but I do like attempts to make more “wholesome” tools that don’t degenerate into machines for cheating on homework and mass-producing slop
- https://priori.chat/ allows you to toggle the “implicit assumptions” in an LLM’s response; though upon playing with it, i consistently preferred Claude’s default choices anyhow. this is definitely the kind of thing i want to exist though.
- https://www.bitsaboutmoney.com/archive/fraud-investigation/ Patrick McKenzie makes the case that we should get more serious about investigating benefits fraud, using the tools of the finance industry

sarahconstantin 5 Feb 2026 19:49 UTC
3 points
0
on: sarahconstantin’s Shortform
links 2/5/26: https://roamresearch.com/#/app/srcpublic/page/02-05-2026
- https://paulgraham.com/mean.html obviously everything Paul Graham said during his tenure at YC was more about what he wanted founders to be like than what they necesssarily were like...but i do think i agree, actually, that mean people fail. at least, I agree with the weaker claim that the most overtly nasty people are usually jobless and socially isolated. successful people may be ruthless, but almost all of them have manners.
- https://www.anthropic.com/news/claude-is-a-space-to-think I like that Anthropic is branding itself as the less-skeezy AI. i want to use products that have a “we’re the Good Guys, we keep it classy” brand identity, even if obviously there are some compromises made for the sake of profit. a company culture of “we don’t do that lowest-common-denominator shit, we have Honor” is obviously toothless from an incentive perspective but can in fact hold out for some time and provide a better-quality customer experience.
- https://en.wikipedia.org/wiki/Assassination_of_Shinzo_Abe somehow i didn’t realize this happened. wild story.
- https://www.construction-physics.com/p/on-technologies-vs-commodities Brian Potter finds that technologies vs commodities are more of a continuum
- https://arxiv.org/abs/2503.21602 scaffolding for text-to-SQL
- https://www.metaculus.com/questions/41139/china-attacksblockades-taiwan-in-2026/ why is this so flat?
- https://www.bbc.com/news/articles/cevz13xkxpro (from Sept 2025) we are finally beating Huntington’s

Biotech Startup Stats

sarahconstantin5 Feb 2026 18:40 UTC

21 points

0 comments4 min readLW link

(sarahconstantin.substack.com)

sarahconstantin 27 Jan 2026 17:46 UTC
14 points
0
on: sarahconstantin’s Shortform
links 1/27/26: https://roamresearch.com/#/app/srcpublic/page/01-27-2026
- https://en.wikipedia.org/wiki/Insurrection_Act_of_1807 last invoked in the 1990s LA riots
- https://www.sluggish.xyz/p/are-we-all-just-having-understandable
  - one annoying thing about anti-psychiatry people is that they equivocate between “the person labeled insane is behaving in a valid fashion, and deserves acceptance/respect rather than pathologization” and “the person labeled insane is causing their own problems and ought to stop; they deserve criticism rather than therapeutic treatment”
  - either of these can, imo, sometimes be true, but it’s weird when someone is trying to pose as both the defender of the misunderstood and the teller of hard truths to malingerers.
- https://www.darioamodei.com/essay/the-adolescence-of-technology not a lot of news in here; Dario Amodei lays out “Anthropic-ism”, the summary of their research & policy programs of the past several years.
  - he and Paul Christiano deserve credit as the architects and prophets of the age we’re living in. things did, after all, turn out their way. i was always critical of their theories as basically blurring boundaries between “real intelligence” and the thing LLMs do, but in their defense, that boundary has turned out to be blurrier than I expected.
  - i think i might be at least sympathetic to Anthropic-ism. it turns out that a world with these kinds of gradually-and-quickly-improving AI tools is pretty nifty, and also involves all kinds of novel risks that are short of the “everyone dies instantly” scenario but are still quite worrying. sure, if they had listened to Eliezer they could simply have not chosen to go down this path, and there’s more than a hint of vainglory in the motivations of people who want to build powerful AI, but all in all I’d rather the 2020s had these tools than not. it’s not like the “end of the end of history” wouldn’t have happened anyway, and at least this way we have some powerful opportunities as well as depressing developments.
- https://oll.libertyfund.org/titles/bagehot-lombard-street-a-description-of-the-money-market 1873 classic about central banking
- https://steve-yegge.medium.com/welcome-to-gas-town-4f25ee16dd04 this sounds quite complicated but interesting. multi-agent management system.
- https://marginalrevolution.com/marginalrevolution/2026/01/a-more-intelligent-comment-than-most-of-the-emotional-reactions-we-are-seeing.html why do comments have to be “intelligent” per se? you can just say the obvious thing that expresses your values! simple things can be correct.
- https://epirium.com/wp-content/uploads/2025/11/GSA2025_Poster_Epirium_MF-300_Phase-1.pdf that sure is a drug for sarcopenia in phase 1 trials
- https://www.axios.com/2026/01/24/ice-cbp-shooting-minneapolis-democrats-reaction imagine, congress actually putting checks on the executive branch!
- https://writing.antonleicht.me/p/how-ai-safety-is-getting-middle-powers too much inside baseball for me, but probably a useful pointer to what experts think
- https://corticallabs.com/ sounds like a joke? you do not want to do any computation on neurons, they are slow and fragile. (you might want to run brain-inspired algorithms, but on semiconductors!)
- https://centuryofbio.com/p/sid extreme personalized medicine approach to cancer—it worked! after the standard of care failed, single-cell sequencing exposed a fibroblast-like phenotype, and an anti-fibroblast radiotherapy shrank the tumor enough to be operable.
- https://www.sam-rodriques.com/post/the-humanity-project this sounds entirely desirable but insanely expensive. sure, i’d like this, and a pony too. and world peace. can you just do that these days? i’d be delighted if so.
- https://defenseanalyses.org/work/the-maha-pentagon/ ehhh like so much DARC stuff it’s hard to tell if they’re serious. biohacking for soldiers?
  - what I’d really like to see in the health/military space is to take advantage of the DOD’s truly massive biobank resources & make more data accessible to researchers from modern omics techniques.
- https://en.wikipedia.org/wiki/Hanfu_Movement historical costume is big in China.
- https://www.humanlayer.dev/blog/writing-a-good-claude-md keep it simple, stupid (one perspective)

sarahconstantin 21 Jan 2026 18:20 UTC
5 points
0
on: sarahconstantin’s Shortform
links 1/21/26: https://roamresearch.com/#/app/srcpublic/page/01-21-2026
- safety of BPC-157:
  - https://www.sciencedirect.com/science/article/abs/pii/S027323002030091X good results in animals
  - https://clinicaltrials.gov/study/NCT02637284 Phase I trial never published results
- efficacy of BPC-157:
- Shaun Tan’s art: https://www.npr.org/2022/11/03/1131598819/shaun-tan-creature-the-lost-thing-the-arrival
- https://merge.io/blog Merge Labs, the successor to Sumner Norman’s Forest Neurotech. Sam Altman is involved.
  - “entirely new technologies that connect with neurons using molecules instead of electrodes, transmit and receive information using deep-reaching modalities like ultrasound, and avoid implants into brain tissue.”

sarahconstantin

FW26 Color Stats

Things that Go Boom

Get­ting Back To It

Biotech Startup Stats

Getting Back To It