Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 19/09/2026Reinforcement Learning: Theory and Algorithms rltheorybook.github.io/rltheorybook... (Theory behind RL techniques used in RLHF) Theoretical Perspectives on Knowledge Distillation: A Review wires.onlinelibrary.wiley.com/doi/10.1002/... (Introduction to model distillation) 5/5 011
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 19/09/2026Reinforcement Learning from Human Feedback rlhfbook.com (Engineering aspects of fine-tuning models to objectives from human feedback or verifiable rewards) 4/5rlhfbook.comRLHF Book: Reinforcement Learning from Human Feedback and LLM Post-TrainingA free online book and course on RLHF, preference tuning, reward models, RLVR, and post-training language models. 121
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 19/09/2026The Emerging Market for Intelligence: How Firms Buy and Sell AI www.aeaweb.org/articles?id=... (Documents competitiveness of open weight models) 3/5aeaweb.orgThe Emerging Market for Intelligence: How Firms Buy and Sell AI(Summer 2026) - We describe the emerging business-to-business market for large language model (LLM) inference and document key empirical patterns in its supply, pricing, and dynamics, using data from… 111
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 19/09/2026This creates opportunities for local / democratic control: - Take open-weight foundation models - Fine tune for local / democratically chosen objectives using RL. Here some useful reading recs: 2/5 111
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 19/09/2026Thread: Key developments in AI / LLMs 1. reinforcement learning (RL, from human feedback or verifiable rewards), 2. open weight models; some based on distillation of commercial models, not far behind commercial alternatives. 1/5aeaweb.orgThe Emerging Market for Intelligence: How Firms Buy and Sell AI(Summer 2026) - We describe the emerging business-to-business market for large language model (LLM) inference and document key empirical patterns in its supply, pricing, and dynamics, using data from… 162
Maximilian Kasy @maxkasy.bsky.social · 19/09/2026Thread: Key developments in AI / LLMs 1. reinforcement learning (RL, from human feedback or verifiable rewards), 2. open weight models; some based on distillation of commercial models, not far behind commercial alternatives. 1/5aeaweb.orgThe Emerging Market for Intelligence: How Firms Buy and Sell AI(Summer 2026) - We describe the emerging business-to-business market for large language model (LLM) inference and document key empirical patterns in its supply, pricing, and dynamics, using data from… 162
Maximilian Kasy @maxkasy.bsky.social · 21/08/2026Jetzt offiziell verfügbar, in deutscher Übersetzung: "Die Mittel der Prädiktion Wie KI wirklich funktioniert (und wer davon profitiert)" www.suhrkamp.de/buch/maximil... In guter Gesellschaft im aktuellen @suhrkamp.de Katalog :-) 1142
Reposted by Maximilian KasySuhrkamp Verlag @suhrkamp.de · 17/08/2026Unsere Neuerscheinungen im August sind da: shrk.vg/Neuerscheinu... @misik.bsky.social @ohneklippo.bsky.social @carlos-barragan.bsky.social @maxkasy.bsky.social @juliannhofmann.bsky.social @isabelallende.com 34612
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 16/08/2026Heute in der FAZ (@faznet.bsky.social): „Die entscheidende Frage der KI lautet: Wer entscheidet, was optimiert wird?“ www.faz.net/aktuell/wirt...faz.netÖkonom Max Kasy: „Die entscheidende Frage der KI lautet: Wer entscheidet, was optimiert wird?“Der Ökonom Maximilian Kasy will die Künstliche Intelligenz entmystifizieren und die Debatte versachlichen. Er spricht über Mathematik, KI als Machtinstrument – und wieso er nicht glaubt, dass der… 0116
Maximilian Kasy @maxkasy.bsky.social · 16/08/2026Heute in der FAZ (@faznet.bsky.social): „Die entscheidende Frage der KI lautet: Wer entscheidet, was optimiert wird?“ www.faz.net/aktuell/wirt...faz.netÖkonom Max Kasy: „Die entscheidende Frage der KI lautet: Wer entscheidet, was optimiert wird?“Der Ökonom Maximilian Kasy will die Künstliche Intelligenz entmystifizieren und die Debatte versachlichen. Er spricht über Mathematik, KI als Machtinstrument – und wieso er nicht glaubt, dass der… 0116
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 02/08/2026Aujourd'hui dans L'Echo (@lechobe, Belgique): www.lecho.be/opinions/gen... 031
Maximilian Kasy @maxkasy.bsky.social · 02/08/2026Aujourd'hui dans L'Echo (@lechobe, Belgique): www.lecho.be/opinions/gen... 031
Maximilian Kasy @maxkasy.bsky.social · 24/07/2026Great interview with @triofrancos.bsky.social on the geopolitics of the renewable energy transition:newleftreview.orgThea Riofrancos, The New World Climate Order, NLR 159, May–June 2026An interview about the new parameters of climate politics, with the breakdown of global consensus. As the EU re-opens coal-fired plants and the US shuts down wind farms, China gouges the earth to… 11510
Reposted by Maximilian KasyQuinn Slobodian @quinnslobodian.com · 14/07/2026Winning at all levels, growth rate, foreign policy, immigration policy, energy policy, list is long 9561137
Reposted by Maximilian KasyNBER @nber.org · 07/07/2026A guide for economists writing a literature review who want to summarize evidence, incorporate covariates, and adjust for selectivity, from Peter Ganong, Avik Garg, and Maximilian Kasy www.nber.org/papers/w35403 1165
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 27/06/2026New paper with @p_ganong & Avik Garg: From literature review to effect prediction for a new context - allowing for effect heterogeneity & publication bias. maxkasy.github.io/home/files/p... 1214
Maximilian Kasy @maxkasy.bsky.social · 27/06/2026Reading rec: "Political Aspects of Full Employment" Intriguing 1943 essay by Michał Kalecki mronline.org/2010/05/22/p... Reasons of business opposition to full employment: It would 1/2mronline.orgPolitical Aspects of Full Employment - MR OnlineThis essay was first published in Political Quarterly in 1943; it is reproduced here for non-profit educational purposes. A shorter version of this essay was published in The Last Phase in the… 171
Maximilian Kasy @maxkasy.bsky.social · 27/06/2026New paper with @p_ganong & Avik Garg: From literature review to effect prediction for a new context - allowing for effect heterogeneity & publication bias. maxkasy.github.io/home/files/p... 1214
Reposted by Maximilian KasySoziopolis @soziopolis.bsky.social · 26/06/2026In „The Means of Prediction” wirft @maxkasy.bsky.social einen kritischen Blick auf Machtfragen rund um KI. @sberg.bsky.social und @annkoster.bsky.social haben das Buch gelesen und setzen es in ihrem Literaturessay zu anderen Debattenbeiträgen ins Verhältnis. www.soziopolis.de/its-the-obje... 0246
Reposted by Maximilian KasyCORE Econ @coreecon.bsky.social · 23/06/2026Chapters 1–4 out now! Understanding our Economy, CORE's textbook for beginners and students at two-year and four-year colleges and universities in North America, is online. Covers the environment, climate, inequality, cost of living & more. Read it free: books.core-econ.org/understandin... 0116
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 20/06/2026Why does Elon Musk stoke racist violence and pogroms, and what are the parallels to 19th century Russia? Thoughtful essay worth reading: giftarticle.ft.com/giftarticle/... Big Tech is stoking unrest in the UK. Why?giftarticle.ft.com 041
Maximilian Kasy @maxkasy.bsky.social · 20/06/2026Why does Elon Musk stoke racist violence and pogroms, and what are the parallels to 19th century Russia? Thoughtful essay worth reading: giftarticle.ft.com/giftarticle/... Big Tech is stoking unrest in the UK. Why?giftarticle.ft.com 041
Reposted by Maximilian KasyTed Underwood @tedunderwood.com · 12/06/2026So, naive question about these campuses that pay OpenAI $13mil/yr. Why don’t they pour that money into a cluster running the strongest available open-weight model, with API and interactive options. And if the answer is “overhead,” why aren’t we collaborating? 5549872
Maximilian Kasy @maxkasy.bsky.social · 29/05/2026www.diepresse.com/26721984/wie...diepresse.comWie KI Ihre persönlichen Daten nutzen kann, um Ihren Nachbarn zu...Die Zukunft der KI wird nicht von klügeren Algorithmen oder schnelleren Chips abhängen. Sie wird davon abhängen, wer die Daten kontrolliert. 042
Reposted by Maximilian Kasyundercoverhist.bsky.social @undercoverhist.bsky.social · 27/05/20261/n Verena Halsmayer giving the Blanqui lecture on the history of the Solow model, based on her wonderful book www.cambridge.org/core/books/m... #eshethes2026 Key insights: 24019
Reposted by Maximilian Kasyryan cooper @ryanlcooper.com · 22/05/2026"Meetings are held, emails are sent, phone calls are made, and something like 520,000 children die of malnutrition, diarrhea, tuberculosis, malaria, and HIV." prospect.org/2026/05/22/w...prospect.orgWho Died When Elon Musk Killed USAID? - The American ProspectNicholas Enrich’s new memoir details the human toll of the agency’s dismantling. 3725501293
Reposted by Maximilian Kasydoew.at @doew.at · 13/05/2026Das lange Warten hat ein Ende: 30 Jahre nach der letzten Auflage ist das neue„Handbuch Rechtsextremismus in Österreich“ seit heute erhältlich. Den Band gibt es in der Buchhandlung Ihres Vertrauens oder direkt beim Verlag des @falter.at: shop.falter.at/handbuch-rec... 16931
Maximilian Kasy @maxkasy.bsky.social · 12/05/2026📆 Only 2 weeks to go until our workshop!!! 📆 "AI and the political economy of information filtering" maxkasy.github.io/home/informa... Open to all; register for partcipation (online/in-person) here: docs.google.com/forms/d/e/1F... 040
Reposted by Maximilian KasyBellingcat @bellingcat.com · 08/05/2026What can flight data tell us about deportation flights, and where are the limits of public information? On Thursday @gillianbrockell.com guides us through the aviation data, public records, and open-source techniques used to follow immigration operations in the skies. discord.gg/bellingcat?e... 321283
Reposted by Maximilian KasyEmilien Schultz @emilienschultz.eurosky.social · 05/05/2026Ce soir, présentation de @maxkasy.bsky.social à @crestumr.bsky.social pour son livre - The means of prediction - ça parle du contrôle démocratique de l’ia (entre autre). Discuté par @eollion.bsky.social qui pose la question : doit on réguler l’IA ou le capitalisme ? 1103
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 03/05/2026Insightful interview on Iran on the eve of the recent invasion: newleftreview.org/issues/ii157...newleftreview.orgErvand Abrahamian, Iran Under Fire, NLR 157, January–February 2026One of modern Iran’s leading historians provides a critical anatomy of the power structures of the Islamic Republic, on the eve of war—and a scathing denunciation of the long-incubated… 031
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 02/05/2026🚨 New paper (with @elizabethlinos.bsky.social and @ProfSanaz) 🚨 "Causal Inference for Social Network Formation" Leveraging random initial ties + panel observation to recover causal effects of indirect ties, degree, and density on who connects. maxkasy.github.io/home/files/p... 0218
Maximilian Kasy @maxkasy.bsky.social · 03/05/2026Insightful interview on Iran on the eve of the recent invasion: newleftreview.org/issues/ii157...newleftreview.orgErvand Abrahamian, Iran Under Fire, NLR 157, January–February 2026One of modern Iran’s leading historians provides a critical anatomy of the power structures of the Islamic Republic, on the eve of war—and a scathing denunciation of the long-incubated… 031
Maximilian Kasy @maxkasy.bsky.social · 02/05/2026🚨 New paper (with @elizabethlinos.bsky.social and @ProfSanaz) 🚨 "Causal Inference for Social Network Formation" Leveraging random initial ties + panel observation to recover causal effects of indirect ties, degree, and density on who connects. maxkasy.github.io/home/files/p... 0218
Maximilian Kasy @maxkasy.bsky.social · 22/04/2026Barry Eichengreen kindly reviewed "The Means of Prediction" in @foreignaffairs.com (May/June 2026) — "a clear and concise explanation of how AI works": 🔗 foreignaffairs.com/reviews/means-prediction-how-ai-really-works-and-who-benefits 070
Reposted by Maximilian KasyDavid Van Dijcke @packlesshepherd.bsky.social · 11/04/2026I coded up an open-source, not-for-profit AI paper reviewer that rivals the performance of @reviewer3com.bsky.social, Refine.ink, and Stanford Agentic Reviewer (according to Gemini 3.1). Costs <$2! Live @ coarse.ink. Plug in paper, @openrouter.bsky.social key, and email. #econsky 34111
Reposted by Maximilian KasyINEQ - Research Institute Economics of Inequality @ WU Vienna @ineq-wu.bsky.social · 07/04/2026🌍 INEQ Research Seminar Who benefits from AI? On April 14, @maxkasy.bsky.social joins our seminar to present his book “The Means of Prediction” – on power, control and inequality in the age of AI. 🗓 April 14, 4:00 pm 🔗 Info & registration: short.wu.ac.at/ineq_va_kasy #INEQ #INEQ_RS 112
Reposted by Maximilian KasyThe New Yorker @newyorker.com · 06/04/2026The economist Arindrajit Dube believes that the answer to fixing America’s wage problem is to empower workers and set mandatory wage standards across industries. www.newyorker.com/news/the-fin...newyorker.comThe Economist Who Wants to Solve America’s Wage ProblemThe economist Arindrajit Dube believes that the answer to fixing America’s wage problem is to empower workers and set mandatory wage standards across industries. 13914
Reposted by Maximilian KasyReuters Institute @reutersinstitute.bsky.social · 03/04/2026What is funding the rapid growth of AI companies? At our recent conference on AI and news, @maxkasy.bsky.social of @oxfordecondept.bsky.social cited the involvement of companies involved in mass surveillance and autonomous weapons systems. buff.ly/r6gCk3h 0147
Reposted by Maximilian KasyTressie McMillan Cottom @tressiemcphd.bsky.social · 02/04/2026Someone sent me this. 196119112957
Maximilian Kasy @maxkasy.bsky.social · 31/03/2026"Cancel-culture" in action: This paper argues that the US governments' blacklisting of research on topics including gender, race, and the environment (for funding) has led to pre-emptive self-censorship in economics working papers. shs.hal.science/halshs-05519...shs.hal.science 030
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 29/03/2026🔥 🌐 Workshop announcement: 🌐🔥 AI and the political economy of information filtering May 25-27, 2026 Workshop website: maxkasy.github.io/home/informa... Registration: docs.google.com/forms/d/e/1F... 052
Maximilian Kasy @maxkasy.bsky.social · 30/03/2026💲💲💲 New version available: 💲💲💲 Basic income and labor supply: Theory and evidence from an RCT in Germany maxkasy.github.io/home/files/p... #basicincome #ubi 22113
Reposted by Maximilian KasyFayad Mulla @fayadmulla.bsky.social · 30/03/2026Stell dir Hilfe vor, die für Jahrzehnte wirkt. Der Verein Energiekollektiv stellt einkommensschwachen Haushalten kostenlos Balkonkraftwerke zur Verfügung --> weniger Stromkosten, mehr Klimaschutz, keine Abhängigkeit von Fossildikaturen. Jetzt spenden: energiekollektiv.at#spendenenergiekollektiv.atDie Balkonrevolution: Nachhaltig helfenWir helfen Menschen mit wenig Geld mit Balkonkraftwerken, damit sie langfristig und nachhaltig vorwärts kommen. 0104
Maximilian Kasy @maxkasy.bsky.social · 29/03/2026🔥 🌐 Workshop announcement: 🌐🔥 AI and the political economy of information filtering May 25-27, 2026 Workshop website: maxkasy.github.io/home/informa... Registration: docs.google.com/forms/d/e/1F... 052
Maximilian Kasy @maxkasy.bsky.social · 29/03/2026giftarticle.ft.com/giftarticle/... Ministers explore triggering break clause in Palantir’s NHS contractgiftarticle.ft.com 020
Reposted by Maximilian KasyMaximilian Kasy @maxkasy.bsky.social · 19/03/2026🔥➗️ New workingpaper: ➗️🔥 "From Cross-Validation to SURE: Asymptotic Risk of Tuned Regularized Estimators" maxkasy.github.io/home/files/p... This was inspired by conversations with Gary Chamberlain 10 years ago, and ended up a heavier lift than anticipated. Any feedback & comments are appreciated!maxkasy.github.io 072
Maximilian Kasy @maxkasy.bsky.social · 26/03/2026Must-read piece on the role of tech and policy in the killing of 180 Iranian school girls by US bombs:theguardian.comAI got the blame for the Iran school bombing. The truth is far more worryingLLMs-gone-rogue dominated coverage, but had nothing to do with the targeting. Instead, it was choices made by human beings, over many years, that gave us this atrocity 1135