kai @kaiq.bsky.social · 30/09/2026I’m a fan of {local publication} but it’s bad practice to let writers write about their subjects with open contempt 100
Reposted by kaiA. Sophia Koepke @askoepke.bsky.social · 17/04/2026Takeaway: Models, like organisms, may perceive the world within their Umwelt (check von Uexküll: en.wikipedia.org/wiki/Jakob_J...). We suspect future evidence will favor von Uexküll over Plato. Different models may learn rich representations of the world, just not the same one. 7/9en.wikipedia.orgJakob Johann von Uexküll - Wikipedia 141
Reposted by kaiJaz @jaz.sh · 27/09/2026Okay so I've built a new Atlas, this time it lets you explore the conversations on Bluesky over the past 7 days graphically. It should update every ~6 hours and help you navigate the information environment that exists on here. I've learned a LOT about what people do here today... atlas.jazco.devatlas.jazco.devAtlasA living map of what Bluesky is talking about: the past week's conversations, grouped into topics and regions and rebuilt every six hours. 1161398414
kai @kaiq.bsky.social · 23/09/2026the more I think about it the more the existence of foosball seems insane 000
kai @kaiq.bsky.social · 22/09/2026There needs to be a word for how a bad technology can still help you by accident, but in a way that is actually a case against the technology 25412
Reposted by kaiThe Last Visible Dog @thelastinstance.bsky.social · 11/08/2026I think this picturing is inadequate to the technical reality, and explaining why might help to clarify why I have different intuitions about what's going on. 2107
kai @kaiq.bsky.social · 13/09/2026is there anywhere with good AI discussion that is not just AI boosters/gearheads (at one extreme), or online leftists who still think AI is stuck in the “stochastic parrot” era (at the other extreme)? T_T 100
Reposted by kaiGoodnews @goodnews.computer · 04/09/2026Take a break from being depressed that everything you love is being destroyed by slop for a moment and let me tell you about the beautiful symbiotic relationship between graffiti writers, train yard workers, and model train enthusiasts. 895483549
Reposted by kaiActual names @actualnames.bsky.social · 11/09/2026Friday Hellions United States, Freedmen's Bureau Hospital and Medical Records, 1865-1872 2275
Reposted by kaiActual names @actualnames.bsky.social · 11/09/2026Ha Zingers Germany, Lutheran Baptisms, Marriages, and Burials, 1500-1971 1322
kai @kaiq.bsky.social · 09/09/2026the sequel to “conflict is not abuse” should be “being cringe is not the same as causing harm” 000
kai @kaiq.bsky.social · 30/08/2026I do find the persona of the careerist tech critic to be contradictory in an interesting way—unlike most careers, it involves building your life’s work on something you hate, while also secretly loving it enough to become intricately familiar with it and wish it were better (e.g. Ed Zitron) 110
Reposted by kaimaya goldberg-safir @roughnotes.bsky.social · 18/08/2026natalie nakase is such a good writer, here's what she just said about communication: “it is high, it is harsh, it is violent, it is direct tonight.” 1345
kai @kaiq.bsky.social · 26/08/2026I wonder if second-gen Asian American art is so computer-centric because that’s the kind of first-gen Asian who was allowed into the US 000
Reposted by kaiMatthew Kirschenbaum @mkirschenbaum.bsky.social · 25/08/2026Machine translation was the original application for the transformer. It’s the use case and benchmarking in AIAYN. 0132
Reposted by kainorvid_studies @norvid-studies.bsky.social · 26/06/2026technical people who were surprised by deep learning success going into what specifically they were surprised by, beyond "hey it works?" [separate thread for readability and linearity] 10286
Reposted by kaiMatthew Kirschenbaum @mkirschenbaum.bsky.social · 31/05/2026In a talk I’ve been giving this spring, I’ve pointed out that “the same authors who are tireless critics of the tendency to anthropomorphize large language models by way of words like ‘know’ and ‘understand’ and ‘think’ seem content to dismiss them based on their inability to ‘mean’ or ‘intend.’” + 1125
Reposted by kaiTed Underwood @tedunderwood.com · 10/06/2026Psst, buddy — fixer 4 is back, and it shipped the wire-up not the cop-out. Dodge a collision and never ride exports: the local badge scoped truthfully. 413318
kai @kaiq.bsky.social · 15/04/2026I do think some friction in technology is good, not because friction is inherently good, but because doing things more slowly actually creates more space for introspection: "am I doing the right thing?" 010
Reposted by kailoscharlos @loscharlos.bsky.social · 18/03/20266 years ago today I went on my last run (pic left) — today I’m 36 and I’ve spent my entire 30’s with #LongCovid struggling to walk up and down the block, or work more than 1-2 hours a day. Heres 4 things I never thought I’d lose at this age: 7320134
Reposted by kaiMaggie Tokuda-Hall @maggietokudahall.bsky.social · 16/03/2026It is genuinely wild that we can claim: Ryan and Zinzi Coogler Alyssa Liu Delroy Lindo Zendaya Marshawn Lynch Mahershala Ali Tom Hanks Angela Davis Chelsea Peretti Boots Riley Maxine Hong Kingston Jack London Fred Korematsu Daveed Diggs Like 812114
kai @kaiq.bsky.social · 09/03/2026“I want to show the world my art” and it’s just some text in a box 000
kai @kaiq.bsky.social · 06/03/2026Objections aside, I largely agree with Rothfeld - I don’t want LLM-written prose or ideation anywhere near my writing, because I want to be part of a community of practice, even if it inconveniences me 010
kai @kaiq.bsky.social · 06/03/2026I’m also seeing a lot of people (including Rothfeld) saying: don’t use AI, use Google! But Google literally is using LLMs for ranking, I’m sure :-/ So that confuses the whole reason to not use AI. Why is Google ok? You should restrict yourself to physical libraries, if AI use is your red line 000
kai @kaiq.bsky.social · 06/03/2026However… I don’t believe that preserving “friction” in life is inherently good! Why would e.g. washing dishes be inherently morally good and human? Now apply the same logic to using AI 000
kai @kaiq.bsky.social · 06/03/2026Feels like the debates about using AI in writing are actually about what reading is for: If reading is for entertainment, then it's fine if what you're reading is AI, as long as it entertains (!) If reading is about making contact with another human mind, as Becca Rothfeld says, then AI is vacuous 000
Reposted by kaiHenry Farrell @himself.bsky.social · 04/03/20269. It can reasonably be objected that the starry-eyed advocates of LLMs are themselves cultists. Even some inside the tribe accept that it is so! But too much Bluesky AI discourse reminds me of vintage Dawkins/Harris atheism - taking up the bad intellectual habits of those they condemn. Finis.asteriskmag.comWhy Are There So Many Rationalist Cults?—AsteriskThere’s a lot to like about the Rationalist community, but they do have a certain tendency to spawn — shall we say — high demand groups. We sent a card-carrying Rat to investigate what’s really going ... 3044028
kai @kaiq.bsky.social · 21/02/2026What’s going on with all the longform essays that send a bemused literary writer to report on some weird phenomenon in the real world with cutting wit… re: gooners, romance novels, and now startup founders?? 000
kai @kaiq.bsky.social · 12/02/2026Finding myself making increasingly gnomic pronouncements because LLMs have trained me to only speak in questions 000
kai @kaiq.bsky.social · 12/02/2026My labor theory is that because all firms have equal access to agents, & agents require high cognitive load to use (see recent HBR study), a smart person using an agent will get 10x more done than just an agent, or a less skilled person using an agent. So to outcompete, you still have to hire people 000
Reposted by kaiAaron Benanav @abenanav.bsky.social · 06/02/2026What might we be able to build in the future that we can't build today? For the @TheIdeasLetter, I wrote about the political economy of technology and technological change in a world beyond capitalism. www.theideasletter.org/essay/a-real...theideasletter.orgA Real Political Economy of Technology - The Ideas LetterTwo technological futures are competing for political and material priority: generative AI and the green transition. Benanav argues that while AI is marketed as a world-reordering breakthrough, its pr... 04015
Reposted by kaiMarco Z @ocramz.bsky.social · 31/01/2026This is not true; I beg people read the full paper and especially the study design. The conclusions mirror my (and many other practitioners') conclusions: if you use AI critically and engage both with the question and the answer, it has a net positive impact on both learning and productivity 96513
Reposted by kaiPavel @spavel.bsky.social · 30/01/2026Yeah this is a classic insight from research into AI and automation in like the 70s that everyone working on LLMs just sort of shrugged about. When the AI's parameters are exceeded and it fails, the untrained human is LEAST qualified to take over. 09712
Reposted by kaihanfinity 🚲 @hanfinity.bsky.social · 30/01/2026I feel like AI coding assistants have the same risks of level 3 or lower vehicle automation, in that they automate the easy and routine stuff which lulls the human into a false sense of security such that the human is not ready to take over when the system encounters something it can't handle. 3706
Reposted by kaiPavel @spavel.bsky.social · 30/01/2026The latest from Anthropic: using Anthropic's products makes you worse at your jobanthropic.comHow AI assistance impacts the formation of coding skillsAnthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems. 15414123
Reposted by kaiAlexandra Lange @langealexandra.bsky.social · 28/01/2026AI coinciding with the white boucle trend in furniture creates many impossible circumstances. The white boucle trend in furniture was itself a product of over-consumption of decor as image because if you thought about it for one second / planned for the longterm you would know it was a bad idea. 051
kai @kaiq.bsky.social · 09/01/2026my New Year’s resolution is actually to post much more and lurk much less, I think this is the ideal use of social media??? 000
kai @kaiq.bsky.social · 16/12/2025ever since I went gluten-free, life has become much more chewy and less toothsome 000
kai @kaiq.bsky.social · 11/12/2025dreamed that I was invited to a partiful where you could only come if you were SuperFriends with the person who invited you (regular friends not allowed) ?? 000
kai @kaiq.bsky.social · 09/12/2025as a software engineer, I have really got to figure out a better apocalypse backup plan. because what am I gonna do when shit hits the fan—vibecode? 010
kai @kaiq.bsky.social · 09/12/2025my silliest thesis is that every avant-grade Asian is connected by at most 2 degrees of separation. not sure what to do with this info 000
Reposted by kaiRob Horning @robhorning.bsky.social · 05/12/2025feel like an ethnography of the people who get paid by the piece to "rank" LLM outputs would also explain a lot about the kind of prose the models are optimized to generate maxread.substack.com/p/will-ai-wr... 23011
kai @kaiq.bsky.social · 20/11/2025the weirdest part of having a corporate job is that I actually don’t feel like a cog in the machine. quite the opposite (they really want you to take the lead, and a lot of teams are built around specific people’s strengths and interests) 100
kai @kaiq.bsky.social · 07/11/2025I swear to god I will stop tweeting about AI so much after I finish this piece! Just working out some ideas 010
kai @kaiq.bsky.social · 07/11/2025It’s a bit strange that people don’t use AI because it makes mistakes, but say that “what makes us human is that we make mistakes”? 000
kai @kaiq.bsky.social · 06/11/2025why did the New Yorker become the main venue for humanistic writing about AI … 000
kai @kaiq.bsky.social · 06/11/2025a very minor complaint in the grand scheme of things, but what I find jarring about leftism is its insistence on totally utopic social relations, its incuriosity about actual social relations, and the particular strain of violence of social relations among leftists 110
