Rachel Leah Childers @donskerclass.bsky.social · 04/10/2026Don't beat yourself up over it. She would have told you but she's super shy.youtu.beNewJeans (뉴진스) 'Super Shy' Official MVYouTube video by HYBE LABELS 130
Rachel Leah Childers @donskerclass.bsky.social · 02/10/2026This is a big part of why I and many people I know are building personalized recommender systems. Not a perfect solution, but a good complement to word-of-mouth. 1120
Rachel Leah Childers @donskerclass.bsky.social · 30/09/2026Derenoncourt has done pioneering work on the distributional effects of policies on wealth doi.org/10.1093/qje/... and wages doi.org/10.1093/qje/... combining detailed micro data, historical context, and effective application of theory, inspiring all of us who work on economic heterogeneity. Brava! 052
Rachel Leah Childers @donskerclass.bsky.social · 30/09/2026What I really want to ask the Keynes bot: 121
Rachel Leah Childers @donskerclass.bsky.social · 26/09/2026I should also mention @brice.green's more social reading list app, papers.club which is more about organizing and sharing your current reading, which is a nice complement to paper search apps. See also lea.ac, which is a Bluesky client with features including highlighting posted papers.papers.clubpapers.club 010
Rachel Leah Childers @donskerclass.bsky.social · 26/09/2026The format was inspired by a similar but more ambitious personalized paper sorter by @ddimmery.com: papers.ddimmery.com Sorting is mostly tf-idf with some similarity comparisons to small local embeddings. Code at github.com/donskerclass... though the hardest part was assembling my own library data.papers.ddimmery.comPapers — new preprints worth readingNew preprints in causal inference, experimental design, statistics and computational social science, ranked by similarity to one researcher's own work. 140
Rachel Leah Childers @donskerclass.bsky.social · 26/09/2026With the rising tide of slop papers, I can no longer mainline the raw nightly arXiv, so I decided to fight slop with slop and vibecode a recommender system based on my library and own interests and papers to sort and select new ones that I would care about. Now live at: donskerclass.github.io/papersdonskerclass.github.ioPapers — Friday 25 September 2026 2123
Reposted by Rachel Leah Childersyaoi gagarin @mel.bzky.team · 21/09/2026FINALLY SOME ACTIONABLE ADVICE 0574
Rachel Leah Childers @donskerclass.bsky.social · 21/09/2026So you are looking for a masculine partner who is not overtly queer but will engage in queer sexual activity and is also an expert in international commerce? Sounds like you want a trade economist. 060
Rachel Leah Childers @donskerclass.bsky.social · 20/09/2026This is an addendum to my previously stated view on the "trans sports issue" which is that sports are about having a fun time with your friends and I like them a lot! 020
Rachel Leah Childers @donskerclass.bsky.social · 20/09/2026My nuanced opinion on trans women in sports is: God I wanna go see a tgirl baseball game so bad! 230
Rachel Leah Childers @donskerclass.bsky.social · 19/09/2026You have a point. That's strong evidence Lola cherry cola and Lola Copacabana are the same woman, years apart. Important info to relay to my gender fluid skateboard crew. 010
Rachel Leah Childers @donskerclass.bsky.social · 19/09/2026Important distinction: Lola showgirl or Lola cherry cola? 100
Rachel Leah Childers @donskerclass.bsky.social · 10/09/2026I used to give a problem set to my students using this guinea pig DAG to teach d-separation, backdoor criterion, etc. It's not nearly so charming when encoded in dagitty and plotted in standard ggdag style - someone clearly needs to add a theme_guinea_pig() 071
Rachel Leah Childers @donskerclass.bsky.social · 07/09/2026Employment data is hard to come by and existing surveys suggest higher rates, indicative of flows, but it's one pathway that can measure these factors. I'm literally always thinking of this @goingdowntv.bsky.social "trans unemployment survey" sketch and wishing we had a trans BLS, CPS, and JOLTS.instagram.comInstagramCreate an account or log in to Instagram - Share what you're into with the people who get you. 100
Rachel Leah Childers @donskerclass.bsky.social · 07/09/2026I thought the results were fairly reassuring; it's better to have data. Given the stories one hears, a 15% callback penalty for being out and (one algorithm's construction of) clocky, and substantially less given factors modifiable with some effort, is not extreme relative to those for other groups. 110
Rachel Leah Childers @donskerclass.bsky.social · 25/08/2026Joyland was great. Somehow one of the best trans movies of the past few years was produced by Malala. 030
Rachel Leah Childers @donskerclass.bsky.social · 20/08/2026See defs 1, 2, 4, and 6 for standard queer usage. See also "fish". The term is most prominent in drag/ballroom culture and has spread from there, so you may see it less in more culturally distant LGBT communities. www.urbandictionary.com/define.php?t...urbandictionary.comUrban Dictionary: cuntycunty: sexy, iconic, epic, never been done before, serving. 020
Rachel Leah Childers @donskerclass.bsky.social · 20/08/2026Important clarifying question: is that Australian K-pop blogger LGBTQIA++? I hesitate because the overlap of K-pop standom and that community is substantial enough that I would honestly be more surprised if the blogger were using the word in the (non-LGBT) Australian way. 100
Rachel Leah Childers @donskerclass.bsky.social · 19/08/2026in re: cunt Is the LGBT Australian usage different from the LGBT US usage, or just more frequent? We have intersectional polysemy to handle here. 110
Rachel Leah Childers @donskerclass.bsky.social · 16/08/2026Also re: macrofounded IO, the firm dynamics literature is basically that. Heterogeneous firms models a la Hopenhayn, patent race models of innovation, Haltiwanger style data work, and substantive questions re: labor share/productivity/trade. It ends up in macro venues because IO people don't like itafinetheorem.wordpress.comModels of Innovation I: The Patent RaceI’ve been going through some old literature on innovation again as part of a current project, so I figured I ought put up a little review of this literature. I’ll cover five strands: th… 041
Rachel Leah Childers @donskerclass.bsky.social · 16/08/2026I think the high stakes and compensation litigation role of IO has made it extremely specialized in a few subareas of what it could be. I do think NEIO gave a good measurement toolkit for micro that mitigates the party trick tendencies of price theory, but it's hard to think abstractly with it. 110
Rachel Leah Childers @donskerclass.bsky.social · 16/08/2026Aggregation, accounting identities, intertemporal budget constraints, all have a respectable place. Price theory is a fun cocktail party trick, but addition is more fundamental than calculus. 140
Rachel Leah Childers @donskerclass.bsky.social · 07/08/2026A nice tutorial on modern gradient-based MCMC samplers highlighting the link between continuous time theoretical analysis and issues that matter to practitioners. This would have been ideal background for understanding modern developments in MCMC I saw at math.ethz.ch/fim/activiti...arxiv.orgFrom Continuous Dynamics to Practical Gradient-Based SamplersGradient-based Markov chain Monte Carlo methods are often introduced as a catalog of algorithms: Hamiltonian Monte Carlo (HMC), the Metropolis-adjusted Langevin algorithm (MALA), the No-U-Turn Sampler... 0204
Rachel Leah Childers @donskerclass.bsky.social · 06/08/2026It seems a bit closer to one of the solution concepts where agents treat a future self as a separate agent (CF below) then play an imperfect info game via Bayes (perfect? One of those sequential refinements) eq'm. Maybe with learning on top? They're at least thinking carefully about info sets. 000
Rachel Leah Childers @donskerclass.bsky.social · 05/08/2026I think it's a metaphor about the tarot readings we got from a girl at the bar over a shot of malort at 1 am, which seemed more meaningful at the time than the day after. 030
Rachel Leah Childers @donskerclass.bsky.social · 05/08/2026"The Abyss" by the way, née The Abyssinian, is the name of the combination queer bar/performance venue/Ethiopian restaurant Mel took me out to in West Philadelphia. Recommended. Get the sambusas. 351
Reposted by Rachel Leah ChildersRachel Leah Childers @donskerclass.bsky.social · 03/08/2026Due to a travel mishap, I have one night in Philadelphia starting now. What should I do in (almost) paradise? 411
Rachel Leah Childers @donskerclass.bsky.social · 03/08/2026Due to a travel mishap, I have one night in Philadelphia starting now. What should I do in (almost) paradise? 411
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026Thanks, I'm testing out Nutpie now! It's simple enough to test because there's already a Julia implementation. Blackjax is already the sampling backend for our Python version, though I know it has a lot more options to explore than I've tried so far. 010
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026Cruel of you to post this right after I deleted my long thread about what the closet was like for me, just because "employers" can be touchy about concepts like "suicidal ideation". TBH though it was largely just a much less funny version of this thread: 120
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026"How many selves are there? At least three. What are they? Don’t play games with me kid." -Rafe on a game-theoretic model of the closet (among other things) by Benabou and Tirole. 181
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026Thanks! GPU stuff probably won't make it into the formal comparisons in the paper since we would need to rewrite a lot of code, but it seems like the way of the future so is likely to be filed under "future work". Logarithmic scaling is pretty tempting. 000
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026I mean, basically no because the whole JAX implementation was a whim by my coauthor with a lot of Claude credits that we're not putting in the paper. I know for a fact the algorithm is subpar, but we would need to implement some exotic linear algebra primitives not currently in JAX to fix it. 010
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026That is definitely worth trying. We find for small models we get a big advantage out of a dense mass matrix, but the n^2 matrix multiply overwhelms that for bigger ones, where diagonal is OK but slows mixing. A better diagonal mass matrix at least has a chance of helping in that case. 010
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026It's (reportedly) great for parallel algorithms on GPUs but not optimized for sequential algorithms on CPUs, where Julia shines. But also, it's not fair in this case since there are algorithmic differences and optimizations in the Julia code while we haven't tried to optimize our JAX implementation. 010
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026This is exactly the kind of thing I was looking for, thanks! I will try out the code and read the paper, which seems to have a comprehensive and up-to-date set of comparisons. 110
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026We have differentiable likelihoods, so gradient-based anything is an option. Code is in Julia but could be black boxed, and also a vibe-coded JAX version for Python though it's several times slower. Applications include a mix of uni- and multi-modal posteriors. 210
Rachel Leah Childers @donskerclass.bsky.social · 31/07/2026OK, it's time to fortify our set of baselines for an MCMC paper. What's the latest in fast/reliable samplers for nonlinear state space models? We have NUTS/HMC/SMC^2/a few variants of Particle Metropolis. There have to be some new update schemes or maybe GPU options worth trying. 560
Rachel Leah Childers @donskerclass.bsky.social · 25/07/2026I guess that means I'm a cis woman now. 🙃 Remember me when I'm gone (actually able to hold down stable employment and housing). 110
Rachel Leah Childers @donskerclass.bsky.social · 25/07/2026OMG, someone found Yu Deng's Xiaohongshu post about the win. I love... PDE 😭 030
Rachel Leah Childers @donskerclass.bsky.social · 24/07/2026No comment on the phrasing, but there do appear to be a class of holders of mostly illiquid assets with high marginal propensity to consume out of current temporary income who appear not to be able to smooth over income fluctuations. Kaplan Violante and Weidner call them "the wealthy hand to mouth."brookings.edu 041
Reposted by Rachel Leah ChildersDr Yuchen Yang @dr-yang.bsky.social · 24/07/2026Yuri forums even made it to his dissertation acknowledgement #girlslove #yuri #anime 021
Rachel Leah Childers @donskerclass.bsky.social · 23/07/2026Inspirational: one of the new Fields Medalists is a himedanshi. 26122
Rachel Leah Childers @donskerclass.bsky.social · 23/07/2026This problem is called "coreset selection" in ML. Methods generally depend on choice of target quantity, but presumably if you are willing to select survey measures you can choose those. Some randomly chosen lit reviews: arxiv.org/abs/1703.06476 arxiv.org/abs/2505.17799arxiv.orgA Coreset Selection of Coreset Selection Literature: Introduction and Recent AdvancesCoreset selection targets the challenge of finding a small, representative subset of a large dataset that preserves essential patterns for effective machine learning. Although several surveys have exa... 250