@FatTempo@the_transit_guy Would suggest you visit Singapore - it's a welfare state, literally currently making trains free during certain hours. Mandated hawker centers like every other block. Gov builds most of the homes. These are literally Mamdani policies. Just also happens to be a 1 party system
@OlivierFon @Ranicket@bernhardsson This has always struck me as a non-natural equilibrium. My instinct is that generally in low friction economies this sort of software would be externally produced and modular like western B2B software. Do you feel this is just a wrong intuition or is there a friction somewhere?
@slatestarcodex@dwarkesh_sp FWIW I do think brute forcing edges across sets of nodes of size ~10,000 is one of the stand out applications of LLMs that is a very obvious LLM-shaped hole that exists today and can often be immediately exploited
@bennpeifert FWIW the interesting thing about LLMs for financial services SaaS is really the generation of novel horizontal datasets that are useful only at scale
@quantian1 This is 100% correct but a non trivial amount of the benefit comes from actually enhancing (not just isolating) idio generation through improving trade behavior
@omarsar0@levi_malott Markdown is great if you're just working with text but if transcribing a pdf containing more difficult tables html is significantly more accurate. The issue is then html uses more tokens so often pdf -> MD with table placeholders + pdf -> tables with html works really well
@francoisfleuret Unsatisfyingly, it's largely about the things you believe should be the case about the macro state you're equivalencing microstates into. The more you think there are legible dynamics at the macro state level, the more you're doing physics
@0xfdf@KrisAbdelmessih@ByrneHobart FWIW I haven't seen a good thematic risk model ever used in practice but the directionality of themes are wildly important. Being balanced across effective risk on and risk off themes is wildly important even if every theme is constructed to be very neutral
@0xfdf@macrocephalopod Empirically the majority of stock pickers generate more idio RoGMV in their longs than their shorts - lots of pressures pushing in that direction, most apparent in trade behavior but also just a well known, easily observable fact
@bucketshopcap "There are so many smart analysts covering semis" means "Fundamental insights are commoditized in semis are not predictive of price going forward" which is equivalent to the insight that forward vol is more opaque and less tethered to analysis
@systematicls Any premia associated with a factor is extremely secondary to the co-movement of the underlying securities. Factors emerge as a function of similar players doing similar things at similar times – that’s 100% of the explanation
@aanalystbro@0xfdf This is viable via SMA - there's no way the existing MLP SMAs don't allow for MLP to concurrently trade in the same account as the subadvisor - certainly not a frequent occurrence but in the case of liquidation very doable. Day to day hedging can be done at the central risk book