@nikicaga Also worth noting that with net registration shifts in FL (mostly driven by migration and demographic turnover), the R margin *would* have actually expanded significantly from 2024 to 2026 if 2024-like turnout patterns had held.
@Socdem_Michael I think this is fine; if the actual mean estimates are for the Dems to have low to mid single digit leads in these contests, statistically you should get results like these sometimes with normal sampling errors. Sign that they’re not herding/fudging more aggressively.
@mattyglesias You make forecasts of a whole bunch of different elections and then over time we see how well the percentages you assigned to everything matched the % of the time it actually happened. Old 538 had some good viz on this
@bdquinn I think that the polls which show Talarico up with electorates whose recalled vote matches 2024 are seeing nonresponse bias within the segment with a Trump 2024 recalled vote, but also, the early vote data strongly implies that the electorate will be *much* bluer than 2024’s.
@MattZeitlin NYC 2025 mayoral primary. Early vote data pretty clearly pointed to a much younger and more Commie Corridor-heavy electorate than pollsters were modeling.
@hecubian_devil Is Zohran kind of a counterexample to this? He also certainly makes spectacle for vertical video, but the spectacle is "city government does some tangible thing that makes life better" and he very unusually has gotten *more* popular since getting elected.
@crawlings13 Bond yield is basically how much future money you have to pay to get present money. If present money costs a lot of future money, it’s harder to make big ticket purchases (house, car) or invest in productive activity (needs to pay off more to justify initial spend).
@Rory_Johnston Steelman might be that they’re achieving an effect similar to absolute traffic-choking by forcing a big risk premium into transit costs, while forcing the US to expend scarcr resources protecting crude tankers?
@andrewho03 I think that you’ve pointed to the answer yourself in the past— capability growth clusters heavily in RL-friendly task domains and lags elsewhere, so you get systems that are superhuman at stuff like code gen and construction-based proofs but are limited in other domains.
@hecubian_devil “Stochastic parrot” is true but extremely misleading when the distribution that it’s stochastic over is “the human text corpus, but fine-tuned for giving formally correct outputs in checkable domains.” The latent structure of that probability space contains a ton of information.
@gbrew24 Sincere question, Greg— you’ve affirmed a “deal imminent” narrative every time that there’s been chatter about one since well, April. Has the fact that it’s failed to materialize every time changed your thinking at all?
@hecubian_devil Like, indispensable social thinker on questions like “what is the origin of the state?” or “where does money come from?”, but like zero understanding of how hard coordinating activity over groups of 10,000+ people gets.
@hecubian_devil I think it’s a reasonable conclusion if you’ve only read “Bullshit Jobs”, but not if you have broader exposure. His great strength and weakness was having much more interested in and having a richer understanding of small-n social groups than large-n ones.
@BretDevereaux Somewhat concerningly, the precedent of the expedition against the Houthis suggests that bombing our way to this outcome might… not be feasible
@ECaliberSeven This only makes sense if you’re assuming sectoral displacement instead of mass workforce obsolescence; if AGI happens, we won’t be coachmen after the internal combustion engine; we’ll be horses.
@mattyglesias Well, also, if building more causes prices to go up because the additional agglomeration effects generate so much economic growth… that’s also an extremely desirable outcome.