2025-12-23
the crowing
Mimetic Theory has an interesting tweet on their AI-powered hedge fund as a first step towards creating superintelligence for economic allocation. In a recent interview, Scott Sumner had an analogy of how monetary policy dominates fiscal policy, where if he’s steering a boat, any attempt by his daughter in the passenger seat to turn the wheel can be easily negated by his superior positioning. I’m curious as to what extent a private actor could effectively manage the economy over government-set monetary and fiscal policy; it would require truly master judoka levels of being able to predict and then redirect the movements of your opponents. The other question is to what extent value creation here would be distinct from value capture, if for example, breaking the bank is more personally lucrative than engineering a soft landing, even taking into account that any winnings can be reallocated into even more superintelligent investments.
Alex Imas and Arpit Gupta write about using LLMs to predict the effects of macroeconomic interventions. The discussion on microfoundations reminds me of my current pet theory that the recent breakdown of formal hiring systems is the primary cause behind the divergence between macroeconomic trends and public economic sentiment, with increased reliance on informal systems of hiring explaining many phenomena like decreasing geographic mobility in the United States. But it’s interesting to consider to what extent having deeper and more “realistic” models will improve your predictive abilities, or whether each additional refinement of underlying assumptions just ends up layering exponentially more assumptions each time. Somewhat related, Samir Varna on the rescuing of the Indian economy from the Mahalanobis model.
While on the topic of hiring, Sam Kahn and Noah Smith both discuss the recent Jacob Savage article on systemic discrimination in prestige hiring. The latter includes a tweet thread by Patio11, noting that this is not just happening in media and academia, but also to a lesser extent within tech. This seems to imply that risk-aversion around anti-discrimination laws is not the reason for the breakdown of the formal hiring system, since that should not be compatible with a successful campaign against the hiring of white and Asian men1. As with dating discourse, it is difficult to separate signalling and true statements in hiring discourse, given how hireability is so closely tied to social status, producing incentives to be somewhat less than forthcoming.
Alberto Romero has a good summary of the “kiki” and “bouba” models of “jagged intelligence”. While on the topic of human relevancy: Ivan Zhao on AI for companies and organizations as equivalent to steel for buildings in providing leverage that extends their capabilities at scale; Toby Ord investigates the METR time horizon benchmark, noting that the cost-performance of frontier models for long-horizon tasks is currently more or less the same as that of the best humans (though presumably this will improve, and already has with Chinese models); Byrne Hobart on how AI tutoring has to potential to dramatically increase the production of prodigies, by providing high quality training data to those who previously would not have had the privilege of access to it.
Ben Shindel writes about David Budden’s recent claims of having solved both the Hodge conjecture and Navier-Stokes.
Greta Dieck in the Metropolitan Review on Will There Ever Be Another You, a Patricia Lockwood novel exploring the schizo separation of meaning from words and objects.
Xiq on tracing the originating tweets behind coordinating strains of thought within post-rationalist twitter.
More likely increased candidate volume without equivalent increases in qualification legibility is the primary cause, or perhaps increasing complexity of work which increases the duration and therefore cost of the internal training period. This suggests that better hiring technology is one possible solution. I’m partial to something like decentralized referral via prediction markets, as a method which is also more incentive compatible and, though not eliminating discretion, at least makes it more apparent.
That being said, another possibility is the one mentioned by Sam, that cultural factors cause certain discriminated groups to be unwilling to complain or bring suits against discrimination. Somewhat related, Evolutionary Psychology podcast interviews Sarah Mathew on the interaction of evolved psychology and cultural norms, Lionel Page on moral norms as emerging out of useful social equilibria, and Lorenzo Warby on private property as an example of that. The top rated comment is a question as to whether this is still occurring, but either way, I expect that it will now be much more common for white men to bring suits with such claims in the near future. What’s less clear is if Asian (and particularly East Asian) men also do the same, and if not, whether they are the new means by which hiring targets are primarily achieved.