2026-02-28
beauty & stupid
Statecraft interviews Jeremy Singer on the US Digital Service’s recovery of the botched Free Application for Federal Student Aid rollout. It’s astonishing that in addition to apparently not learning anything from Healthcare.gov, there are indications that they haven’t internalized FAFSA’s lessons either. This actually makes me a lot less confident that USDS approach of deferential guidance is actually an effective approach, since it apparently means they don’t get called in proactively, but only for firefighting. The tech force’s approach of forcible embedding might actually be what’s required, particularly as progress in AI means that one engineer can become the single point of contact for all requirements. Given that the requirements process evolved out of the need to coordinate between many (low-quality) external vendors, if you replace the vendors with agentic swarms, then the process itself becomes increasingly unnecessary. Once outcomes become both more salient and verifiable than processes internally speaking, so long as sufficient political cover is provided, then it will be possible to ignore the latter1.
On a somewhat related note, the big news of the day is that the US Department of War has indeed decided to impose supply chain risk designation against Anthropic, which potentially affects their ability to operate on AWS2. ChinaTalk has an excellent podcast on this, where they correctly note that the issue at stake isn’t really about specific policy, but the question of external oversight. After all, the OpenAI contract is functionally the same terms, the only difference being the degree of safeguards that Anthropic feels necessary to ensure that they are actually being followed. It seems to me that the fundamental issue with the current administration is that they view themselves as great men of history who are escaping the shackles of the system, not realizing that the truly great men of history are those who can create and then integrate themselves within a system which goes on to achieve great things. The Defense Production Act is so broad and discretionary because, as a gentleman’s agreement, it expects the government to act within reasonable norms. When the norms are violated, the system reacts by imposing hard rules and processes. When the rules are violated, the system demands oversight3. If oversight is overthrown, then everybody runs away, and you end up as the king of nothing. No one accepts such an imbalanced unilateral relationship, not even within hierarchical relationships like master and servant or parent and child. In this circumstance, the government is actually the child: if they are not competent to build it, then they probably also not competent to know when to use it, since won’t know what they don’t know. It’s correct to want to eliminate processes and focus on outcomes as a means to improve speed of delivery, but people only have the confidence to grant discretion if you accept their oversight and demonstrate your capacity to behave within acceptable norms.
Ajeya Cotra on her faith in the arc of the universe bending towards the good. It does seem like everything which is occurring in AI is indicating a strong relationship between capabilities and the notion of goodness. Though personally I see this more as evidence for morality as being culturally evolved than proof of Platonism. Which means the reward does not just fall out of good intentions; we need to keep thinking about what leads to good outcomes.
Jerusalem Demsas on the measurably large impacts that housing regulations has on elevated housing prices.
Vishal Prasad on Asian-American art being bad. Personally, I see this is a confusion conflation of authorial identity and the identitarian form. For the latter, it’s indeed because the content is ill-suited for the form, drawing as it is from the Irish and African-American traditions of oppressed alienation, or perhaps the Jewish-American one of being really horny. No one can identify with a female Phillip Roth. Actually, if Portnoy’s Complaint came out today, it probably also wouldn’t be very well received.
Peter McLaughlin on the cultural perceptions of Japan as partially created by Lafcadio Hearn.
Abundance and Growth linkthread.
Somewhat related, Ruxandra Teslo in Asimov Mag on how “AI Won’t Automatically Accelerate Clinical Trials” (reiterating an earlier post from her blog). I feel like this leaves out the words “in the United States”, because if clinical trials reform don’t take place, the end result probably won’t be one where everyone still keeps ramming their heads against the US system. We might get another one of those cases where policy wonks in the US describe their perfect system, and then it ends up getting implemented somewhere else first.
This is another case where the humanities and the left are totally abdicating their responsibilities by pretending that AI is nothing. There’s Henry Farrell and Max Read, but for the most part this whole situation seems greatly underexamined from that perspective.
Related, Matt Glassman on IEEPA, and how the Supreme Court basically put itself in its current position by stripping out the legislative veto, in doing so transferring the role of oversight from Congress to themselves.