2026-03-09
write once, run melos
Adam Marblestone on Eon Labs' recent announcement of a successful fruit fly brain simulation. With all caveats about lack of proof and possibly exaggerated claims, it seems to me that advances in connectomics have by far the greatest potential for expanding human flourishing while also preserving human agency and optionality: unlocking non-exclusionary options as disparate as mind-uploading, robotic body-replacement, and human-AI fusion. It will be very interesting to see if simulators can figure out how to unlock neuroplasticity, such that uploads are able to create memories and learn new behaviors, not only for potential applications to LLMs, but also because claims of 91% “behavior accuracy” suggest that the upload process produces an amount of damage essentially equivalent to undergoing a severe stroke. (Edit: more details).
Evolutionary Psychology Podcast interview with Amy Boddy on pregnancy and other forms of microchimerism from an evolutionary approach, analyzed from the cellular level. Actually much of mom discourse has become more understandable to me given the framing that there exists a genetic conflict between a mother (who shares 100% of DNA with herself), and her child (who “only” shares 50%). Presumably this is part of the relationship between decreased family size and trends towards more intensive parenting styles, since this conflict is premised on the mother being able to extend her genetics by having other children later.
Cremieux on Hired as a case study of how making salary ranges publically available can ameliorate pay-gaps.
Lionel Page good overview on how cooperation is game-theoretically optimal behavior which naturally emerges out of repeated interactions1.
Eurydice on poorly informed and overblown takes on AI biorisk. It’s particularly annoying when people say things like “AI is as powerful as a nuclear weapon, and Anthropic is stupid if they think they can retain control of it”. Even assuming a completely naive superintelligent AI, anyone can look at Iran to understand that knowing how to build a nuke is not equivalent to actually having one2.
Feyi Fawehinmi and Tobi Lawson finish their chapter-by-chapter critical review of How Africa Works.
Chris Arnade talks about the difference between South Korea and America as a communal versus individualist society, which I wish people wouldn’t do, because it implicitly accepts the Schmittian worldview that liberalism is to blame for American crime and disorder (as described by Phil Magness in The Argument). As far as I can tell, every culture is rule-following, norm-following, individualistic, and ingroup favoring; the primary distinction is that the rules and norms that any particular group follows (and in what areas the individual versus the group maintains supremacy) as a matter of course are not noticed except when they are broken, while the water that others are swimming in tends to be considered notable. In descriptions of East Asian societies, communal is variable used to mean hierarchical (except when drunk), rule-bound (except for the rules that are ignored), norm-following (unsurprising, given they are norms), prosocial (except for their culture-specific antisocial behaviors), and shame-based (as opposed to performance-based, which presumably has nothing to do with the perceptions of others in one’s community). The word is really just far too overloaded for anyone to understand what it really means; it ought to be tabooed. The better framing is what Page notes, that “There are endless ways to cooperate, hence culture matters”. And there is ample room in the liberal framework to understand and manage the concept of negative externalities.
Metaphorically, the bikini is also as powerful as a nuclear weapon, hence it’s name. I haven’t heard anyone suggesting that only Hegseth should be allowed to wear one3. Anyway, Benny Morris in Quillette with a good overview of the Iran War as a conflict expanding to and involving the entire Middle East.
Edit: More seriously, there is a very obvious and simple method which can constrain the eventual dominance of the frontier AI labs and their employees, which is an equivalent to the popular idea of limiting AI by not allowing it to touch the physical world in any way. Essentially, forbid them from investing or engaging in any non-software products, particularly robotics; cap the total number of employees they are allowed to have, limiting their access to physical labor; create a required company ownership structure, preventing access to shell companies as a means of evasion. Essentially, only allow them to exist as a service provider to others. One might think “but that will disincentivize people from working at frontier AI labs!” If so, what does that imply about where the power of AI really comes from?