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://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.
Mythos-style models that create security vulnerabilities
Nitpick: Security vulnerabilities are typically created by mistake; by coders (human or AI) writing buggy code. Mythos is noted for discovering vulnerabilities, not creating them.
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.
Nitpick: Security vulnerabilities are typically created by mistake; by coders (human or AI) writing buggy code. Mythos is noted for discovering vulnerabilities, not creating them.