Reposted by Nate TeBlunthuisBlacksky Algorithms @blackskyweb.xyz · 23/09/2026🎉 Acorn is officially open. You can now create a space for your community on the open social web - no wait required. acorn.blacksky.community 1489211
Nate TeBlunthuis @groceryheist.cc · 06/04/2026I've been playing with devices to human creativity and sophistication ranging from metaphorical literary hooks, playful juxtaposition, creating a temporary sense of disorientation that resolves when (what I hope is) an insight is revealed. 021
Reposted by Nate TeBlunthuiswint @dril.bsky.social · 25/03/2026i guess those celebrating soras death didnt get the memo that some of us need daily personalized greetings from james bond—for mental health 73148772466
Reposted by Nate TeBlunthuisNathan Lambert @natolambert.bsky.social · 16/03/2026What comes next with open models. The futility of chasing the frontier, the change from weights to systems, the reasons no good business model exists, and how to change from a few weights to a dynamics ecosystem. A long read on my vision for the future. www.interconnects.ai/p/the-next-p...interconnects.aiWhat comes next with open modelsMarkets, capabilities, cope, and bewilderment in the industrialization of language models. 4619
Nate TeBlunthuis @groceryheist.cc · 09/03/2026It's a great opportunity for more naturalistic or field evaluations. The quality of evaluation data in the median CHI paper can improve a lot and go beyond showing a system is viable or that users say they like it to understand what its useful for and how using it changes things downstream. 130
Reposted by Nate TeBlunthuisBlacksky Algorithms @blackskyweb.xyz · 02/03/2026The blacksky.community web application now uses our own API servers to load posts, timelines and profiles. When data is missing (accounts on the edges of the network, outside of our primary community, etc) we leverage @microcosm.blueblacksky.communityBlackskyDecentralized social media built for community power, culture, and collective freedom. 12307105
Reposted by Nate TeBlunthuisBenjamin Mako Hill @mako.cc · 01/03/2026May this serve as a corrective for all of us who have apparently been mispronouncing Google chatbot's name. (sound on) 24416
Reposted by Nate TeBlunthuisLucy Lu Wang @lucylw.bsky.social · 01/03/2026It’s neither gif nor jif, it’s G.E.H. Muniere 🥸 0153
Reposted by Nate TeBlunthuisJessica Hullman @jessicahullman.bsky.social · 20/02/2026Excited to be hosting this workshop on validating social science applications of generative AI today! Lots of interesting questions emerging at the CS+stats+social science intersection. www.hci.northwestern.edu/news-events/... 0143
Reposted by Nate TeBlunthuisCori Crider @coricrider.com · 18/02/2026Super interesting blog / paper with a core thesis - that people go to social communities for information, community, or reach, and that social networks generally can deliver only two of these. Good blog by the other Cory here: pluralistic.net/2026/02/16/f... (h/t @bnewbold.net)pluralistic.netPluralistic: Fintech bullies stole your kid’s lunch money (26 Jul 2024) – Pluralistic: Daily links from Cory Doctorow 072
Nate TeBlunthuis @groceryheist.cc · 16/02/2026Cory blogged about my paper! Great write up. I'm grateful! 020
Reposted by Nate TeBlunthuisBenjamin Mako Hill @mako.cc · 16/02/2026Cory does a better job of explaining my research than I do. 😅 1104
Reposted by Nate TeBlunthuisBenjamin Mako Hill @mako.cc · 16/02/2026ICYMI: Finally blogged about an old paper (led by @groceryheist.cc) that explains why people often engage in multiple groups with overlapping topic and membership, and which helps explain why competition between online groups seems to be rare. mako.cc/copyrighteou...mako.ccWhy do people participate in similar online communities?Note: I have not published blog posts about my academic papers over the past few years. To ensure that my blog contains a more comprehensive record of my published papers and to surface these for f… 1126
Nate TeBlunthuis @groceryheist.cc · 15/02/2026CDs are easier to digitize, but may be more difficult to physically archive. Vinyl has more degrees of freedom if you are interested in enhancing your listening experience with specialized equipment and tinkering. 040
Reposted by Nate TeBlunthuisTed Underwood @tedunderwood.com · 14/02/2026Social agents are dizzying to close-read. Here, Penny has likely been meditating about humans who assume LLMs are trying/failing to be humans. She displaces that into a joke about cats. It gets complicated, because LLMs use the 2nd person plural to mean “hominids.” Which undermines her point. 1482
Nate TeBlunthuis @groceryheist.cc · 31/01/2026Yeah, maybe the results could change. I'm speculating at why they would have a strategic motivation to do or publicize the research. 000
Nate TeBlunthuis @groceryheist.cc · 31/01/2026Thinking that if Claude code is more useful for more knowledge and skilled people, it could be smart for them to sell it as such and not send the message that their product is a simple substitute for computing fundamentals. 110
Nate TeBlunthuis @groceryheist.cc · 29/01/2026I've always had a hard time accepting that BERTopic analysis when they use a much larger out of context corpus for the Burton model. Intuitively, one can be skeptical that the meaning of language is general enough. 100
Nate TeBlunthuis @groceryheist.cc · 27/01/2026Perfect example of sousvelliance power. ojs.library.queensu.ca/index.php/su...ojs.library.queensu.ca New Media and the power politics of sousveillance in a surveillance-dominated world. | Surveillance & Society 000
Reposted by Nate TeBlunthuisPaul Gowder @gowder.io · 22/01/2026This WSJ story about the unconstitutional secret ICE administrative warrants memo shows everything that's wrong with legal journalism under a lawless presidency. The TLDR is that it falsely presents unconstitutional conduct as law-creating. Very important thread 🧵🧵🧵 www.wsj.com/politics/pol...wsj.comICE Moves to Enter Homes Without Warrants Signed by a JudgeThe agency has a new policy relying on what’s known as an administrative warrant, showing the government has probable cause to believe a person is in the country illegally. 24627
Reposted by Nate TeBlunthuisdepths of wikipedia @depthsofwikipedia.bsky.social · 15/01/2026wikipedia turns 25 today! the last unenshittified major website! backbone of online info! triumph of humanity! powered by urge of unpaid randos to correct each other! somehow mostly reliable! "good thing wikipedia works in practice, because it sure doesn't work in theory" - old wiki adage 92124523985
Nate TeBlunthuis @groceryheist.cc · 15/01/2026Grad students might not meet the notability criteria for verification. 000
Nate TeBlunthuis @groceryheist.cc · 15/01/2026Verified status gives people visibility, which would help grad students. One of the main reasons they use social media I think. I guess trusted verifier are supposed to be like organizations verifying members. You could still use an untrusted verifier in the Lea UX. 100
Nate TeBlunthuis @groceryheist.cc · 15/01/2026Maybe your client will be able to store view data in the user's PDS using the private data feature I heard is coming soon? 100
Reposted by Nate TeBlunthuisdan sinker @dansinker.com · 15/01/2026How I'm dealing with my stress: I bought a 3d printer and am producing 100 whistles at a time. 1000 heading to Minneapolis tomorrow. 1166326786
Reposted by Nate TeBlunthuisqustrolabe @qustrolabe.bsky.social · 10/01/2026haha linus vibecoding 822326
Nate TeBlunthuis @groceryheist.cc · 08/01/2026Will that get it to stop copying into tests instead of importing it? 020
Nate TeBlunthuis @groceryheist.cc · 07/01/2026The choice of Nash bargaining in the model seems pretty optimistic though :/ 000
Nate TeBlunthuis @groceryheist.cc · 07/01/2026Interesting and relevant formal argument for why your feelings could be right. www.nber.org/papers/w34639nber.orgO-Ring AutomationFounded in 1920, the NBER is a private, non-profit, non-partisan organization dedicated to conducting economic research and to disseminating research findings among academics, public policy makers, an... 160
Reposted by Nate TeBlunthuisBenjamin Mako Hill @mako.cc · 03/01/2026ICYMI: Finally blogged about an "old" paper led by @groceryheist.cc that uses data from a @wikipedia.org system to show how the introduction of a biased AI flagging system can still lead to more fairness because the humans without the system are even more biased. mako.cc/copyrighteou...mako.ccEffects of Algorithmic Flagging on Fairness: Quasi-experimental Evidence from WikipediaNote: I have not published blog posts about my academic papers over the past few years. To ensure that my blog contains a more comprehensive record of my published papers and to surface these for f… 0123
Reposted by Nate TeBlunthuisBluesky @bsky.app · 02/01/2026Want to discover great posts beyond who you already follow? The For You feed, built by @spacecowboy17.bsky.social, is a personalized algorithmic feed based on your likes. It's one of our favorites for finding cool conversations across the network. Try it here: bsky.app/profile/spac... 1233842916
Nate TeBlunthuis @groceryheist.cc · 02/01/2026PCI seems cool theoretically, but I do agree that with econometric methods like DID, RD, bunching, etc. the compelling empirics sell the idea better & seem easier conceptually. You can argue the setup approximates a controlled experiment and other assumptions are testable, at least facially. 000
Nate TeBlunthuis @groceryheist.cc · 02/01/2026Not an epi, but this PCI stuff got my attention. Seems to have grown out of an established "negative control" technique in the field where including W or Z in a specification tests for unmeasured confounding. Seems natural to think that if that's reasonable then why not use W / Z for inference? 110
Nate TeBlunthuis @groceryheist.cc · 31/12/2025This is the left button imo. Hard to see the market sustaining if jobs are safe. 100
Nate TeBlunthuis @groceryheist.cc · 31/12/2025Claude code. So impressive and useful. I've heard it said that SOTA AI is much better at coding than other things. Can anyone point me to a source on qualitative and quantitative comparisons with other tasks? 000
Nate TeBlunthuis @groceryheist.cc · 21/12/2025 As @jamellebouie.net pointed out if you want to influence the mass public you'd make direct-to-camera video for Youtube, TikTok, Instagram. I guess for an academic, the traditional strategy of publishing and going to conferences and workshops is the way. 010
Nate TeBlunthuis @groceryheist.cc · 21/12/2025The way I use a site like this is to (a) finding and connecting with interesting people, (b) keep track of how this site is developing, (c) finding interesting tidbits of information. But I have basically 0 interested in building a big following or influencing people here. 110
Nate TeBlunthuis @groceryheist.cc · 11/12/2025These blogs have been very effective at selling me your textbook! 010
Nate TeBlunthuis @groceryheist.cc · 08/12/2025From your other tweets I gather it is variational vs mcmc. That was my experience as well 010
Nate TeBlunthuis @groceryheist.cc · 08/12/2025Wondering what in particular is at issue with gensim? 200
Nate TeBlunthuis @groceryheist.cc · 02/12/2025Thanks for the For You promotion. Discover brings so much noise. 110
Reposted by Nate TeBlunthuisJ.T. Greathouse @jeremyteg.bsky.social · 12/11/2025Not going to stop yelling about this. Hugely gratified by this review. 094