#  @drmtown Danny Wilf-Townsend Danny Wilf-Townsend posts on X about ai, at least, thomson reuters, generative the most. They currently have [-----] followers and [--] posts still getting attention that total [---] engagements in the last [--] hours. ### Engagements: [---] [#](/creator/twitter::54990276/interactions)  - [--] Week [---------] +16,916% - [--] Month [---------] -93% - [--] Months [----------] +2,283,463% - [--] Year [----------] +94,754% ### Mentions: [--] [#](/creator/twitter::54990276/posts_active)  ### Followers: [-----] [#](/creator/twitter::54990276/followers)  - [--] Month [-----] +0.98% - [--] Months [-----] +0.26% - [--] Year [-----] +0.26% ### CreatorRank: [---------] [#](/creator/twitter::54990276/influencer_rank)  ### Social Influence **Social category influence** [technology brands](/list/technology-brands) 8.33% [stocks](/list/stocks) 5.56% [travel destinations](/list/travel-destinations) 5.56% [countries](/list/countries) 2.78% **Social topic influence** [ai](/topic/ai) 30.56%, [at least](/topic/at-least) 8.33%, [thomson reuters](/topic/thomson-reuters) #45, [generative](/topic/generative) #610, [law](/topic/law) 5.56%, [bloomberg](/topic/bloomberg) #121, [open ai](/topic/open-ai) 5.56%, [georgetown](/topic/georgetown) 5.56%, [op](/topic/op) 5.56%, [glad](/topic/glad) 5.56% **Top accounts mentioned or mentioned by** [@profarbel](/creator/undefined) [@virginialawrev](/creator/undefined) [@stanfordhai](/creator/undefined) [@adamunikowsky](/creator/undefined) [@vauhinivara](/creator/undefined) [@thezvi](/creator/undefined) [@garymarcus](/creator/undefined) [@ft](/creator/undefined) [@brianwolfman](/creator/undefined) [@dwarkeshsp](/creator/undefined) [@binarybits](/creator/undefined) [@markzjia](/creator/undefined) [@peterwildeford](/creator/undefined) **Top assets mentioned** [Thomson Reuters Corporation (TRI)](/topic/thomson-reuters) ### Top Social Posts Top posts by engagements in the last [--] hours "@TheZvi Am I the only one who finds that it hallucinates like a sailor (Or whatever the right metaphor is). I still have plenty of uses for it but in my field (law) it feels like it makes it harder to convince the many AI skeptics when much-touted models make things up left and right" [X Link](https://x.com/drmtown/status/2021054252614119424) 2026-02-10T02:50Z [----] followers, [---] engagements "A periodic update about the frequency and intensity of AI use in legal practice: Thomson Reuters reports that 55% of generative AI users at law firms use it at least daily with 30% multiple times a day:" [X Link](https://x.com/anyuser/status/2020960620561170858) 2026-02-09T20:38Z [----] followers, [--] engagements "Of course that doesn't tell you what proportion of lawyers are "current users." Last spring Bloomberg found that a majority of lawyers "have used" generative AI for work which is at least something but isn't quite the same as "currently use."" [X Link](https://x.com/anyuser/status/2020960622578438524) 2026-02-09T20:38Z [----] followers, [--] engagements "Thomson Reuters link: Bloomberg link: https://aboutblaw.com/bjbL https://www.thomsonreuters.com/content/dam/ewp-m/documents/thomsonreuters/en/pdf/reports/2026-ai-in-professional-services-report.pdf https://aboutblaw.com/bjbL https://www.thomsonreuters.com/content/dam/ewp-m/documents/thomsonreuters/en/pdf/reports/2026-ai-in-professional-services-report.pdf" [X Link](https://x.com/anyuser/status/2020960624700985545) 2026-02-09T20:38Z [----] followers, [--] engagements "This is an important issue and certainly a development worth following but I would hesitate to try and learn lessons or advice for clients until there is at least a transcript or a written ruling with a sense of Judge Rakoff's reasoning. Not much to go on yet. Your AI conversations aren't privileged. Yesterday Judge Jed Rakoff ruled that [--] documents a defendant generated using an AI tool and later shared with his defense attorneys are not protected by attorney-client privilege or work product doctrine. The logic is simple: an AI Your AI conversations aren't privileged. Yesterday Judge Jed" [X Link](https://x.com/drmtown/status/2021957397414281306) 2026-02-12T14:39Z [----] followers, [---] engagements "Looking forward to the philosophers of the future: can Claude create an eval so difficult that even He can't saturate it Soon enough all evals will be saturated we won't be able to make better ones and the only thing left will be vibes and "guardrails" Soon enough all evals will be saturated we won't be able to make better ones and the only thing left will be vibes and "guardrails"" [X Link](https://x.com/anyuser/status/2022048276787474599) 2026-02-12T20:40Z [----] followers, [---] engagements "Another important dynamic here is that even a very small proportion of lawyers misusing LLMs will result in a steady stream of stories about hallucinated cases. But it won't tell you anything about how whether / how LLMs are useful for lawyers who are acting responsibly. @GaryMarcus @FT As a lawyer who uses LLMs every day at work I feel qualified to respond. First hallucinations are no longer a problem. Consistent with the prediction you quoted from [----] GPT-5.x almost never hallucinates. And overall the percentage of inaccurate responses I get from GPT-5.2 @GaryMarcus @FT As a lawyer who" [X Link](https://x.com/drmtown/status/2022343823276515744) 2026-02-13T16:15Z [----] followers, [---] engagements "I haven't spent much time with the new o3 model from OpenAI but it is the first model to get all of the questions right on the informal testing of legal questions that I've been doing for a while when new models come out" [X Link](https://x.com/drmtown/status/1912890255293792570) 2025-04-17T15:25Z [----] followers, [----] engagements "The recent discovery order in Huskey v. State Farm highlights what is likely to be an increasingly common issue in "AI" cases: what specific decision-making tools are we talking about and how are those integrated into all of the other decision processes that a company has" [X Link](https://x.com/drmtown/status/1762208869013406011) 2024-02-26T20:11Z [----] followers, [---] engagements "Excited to say that my article Deterring Unenforceable Terms will be coming out in the @VirginiaLawRev. It builds on the great empirical work on unenforceable terms in recent years focusing on the normative case and policy details for affirmative penalties for contract drafters:" [X Link](https://x.com/drmtown/status/1765737818943529121) 2024-03-07T13:54Z [----] followers, [----] engagements "@brian_wolfman @VirginiaLawRev Thank you It's less ambitious than saying contract should be full of tort dutiesit's arguing for affirmative penalties for mass drafters which could take a few formsstatutory damages a judge-made cause of action possibly ethics sanctions for lawyer drafters etc" [X Link](https://x.com/drmtown/status/1765742054100152503) 2024-03-07T14:11Z [----] followers, [---] engagements "A new study from @StanfordHAI evaluates legal research AI tools. As part of Georgetown Law's AI committee this year I've had lots of conversations with academics and practitioners about these tools and this paper highlights two important points: https://hai.stanford.edu/news/ai-trial-legal-models-hallucinate-1-out-6-queries https://hai.stanford.edu/news/ai-trial-legal-models-hallucinate-1-out-6-queries" [X Link](https://x.com/drmtown/status/1795811898224517601) 2024-05-29T13:38Z [----] followers, [---] engagements "(Things can get more nuanced than just "affirm/reverse" and Claude is much better at capturing that nuance so it is actually much better than an auto-reverse program. But the baseline point is still important.) 4/" [X Link](https://x.com/drmtown/status/1808557499806855383) 2024-07-03T17:44Z [----] followers, [---] engagements "But more important is the question of qualityis Claude as good as a SCOTUS clerk Here I think a ton of Claude's answers' quality comes from the briefs: SC briefs are typically excellent clear and persuasive with material for Claude to use to generate high-quality responses" [X Link](https://x.com/drmtown/status/1808557809707200864) 2024-07-03T17:45Z [----] followers, [---] engagements "@AdamUnikowsky That is just pretty different from what judges and lawyers do much more often which involves not making up legal standards but determining the best answer for what existing standard applies in a given context and how. And for that a good law clerk needs to be able to look /10" [X Link](https://x.com/drmtown/status/1808558821859864577) 2024-07-03T17:49Z [----] followers, [---] engagements "@AdamUnikowsky beyond the briefs and engage in legal research and reasoning without the assistance of SCOTUS lawyers clearly summarizing and writing things up. And on that front we know LLMs have a *lot* of problems. /11 https://x.com/drmtown/status/1795811898224517601 A new study from @StanfordHAI evaluates legal research AI tools. As part of Georgetown Law's AI committee this year I've had lots of conversations with academics and practitioners about these tools and this paper highlights two important points: https://t.co/LtDPOAAYds https://x.com/drmtown/status/1795811898224517601 A new" [X Link](https://x.com/drmtown/status/1808558946954924046) 2024-07-03T17:50Z [----] followers, [----] engagements "I enjoyed reading "Seeing Like a State" and learned from it but I think everyone who likes it would also benefit from reading Paul Seabright's critical reviewa strong candidate for the best book review I've ever read: https://www.lrb.co.uk/the-paper/v21/n11/paul-seabright/the-aestheticising-vice https://www.lrb.co.uk/the-paper/v21/n11/paul-seabright/the-aestheticising-vice" [X Link](https://x.com/drmtown/status/1815390918297661522) 2024-07-22T14:18Z [----] followers, 32.7K engagements "Model deletion aka algorithmic disgorgement is a remedy the FTC developed in recent years that has received a fair amount of praise with other litigants now seeking it too in AI / privacy / IP contexts. But it has a flaw: the potential to be grossly disproportionate (🧵)" [X Link](https://x.com/drmtown/status/1828066292277911682) 2024-08-26T13:45Z [----] followers, [---] engagements "Where there's been a legal violation involving data model deletion requires the destruction not only of the data but also of tools trained on that data. Here is the NYT requesting it in its lawsuit v. OpenAI and the FTC requiring it in a facial recognition case (Rite Aid) 2/" [X Link](https://x.com/drmtown/status/1828066476705685697) 2024-08-26T13:46Z [----] followers, [---] engagements "Charming to find out that the most popular 20th C. poet in Japan was also the translator of the (extremely popular in Japan) Peanuts comic strip. Theres no perfect analogue but imagine if Robert Frost had translated all of the Asterix or Tintin comics https://www.nytimes.com/2024/12/06/books/shuntaro-tanikawa-dead.htmlsmid=nytcore-ios-share&referringSource=articleShare https://www.nytimes.com/2024/12/06/books/shuntaro-tanikawa-dead.htmlsmid=nytcore-ios-share&referringSource=articleShare" [X Link](https://x.com/drmtown/status/1867163672210231382) 2024-12-12T11:04Z [----] followers, [--] engagements "@ProfArbel This is plausible but I'm not sure it is my initial takeaway. It seems like another read might just be about market segmentation: o3 was way better than 4o but a very surprising (to me at least) proportion of people never clicked through to it and just defaulted to 4o. So " [X Link](https://x.com/drmtown/status/1955671061242212545) 2025-08-13T16:41Z [----] followers, [--] engagements "@ProfArbel .those people who apparently are a supermajority of ChatGPT users are likely to experience [--] as a significant step up because they didn't (or rarely) encounter reasoning models in their previous uses" [X Link](https://x.com/drmtown/status/1955671405334569010) 2025-08-13T16:42Z [----] followers, [--] engagements "@ProfArbel So I have been wondering if the main updating here to my thinking should maybe be "OpenAI cares more about the median consumer" than I had thought (vs. caring about consumers at the frontier of capabilities). As opposed to updating on e.g. talent diaspora" [X Link](https://x.com/drmtown/status/1955672407567638784) 2025-08-13T16:46Z [----] followers, [--] engagements "@ProfArbel Yes definitely agree that these aren't mutually exclusive. Re the mad scientist frontier lab one thing I'll be looking out for is what ends up happening with the more intelligent model they claim to have internallywhen/how is that released how does it end up doing etc" [X Link](https://x.com/drmtown/status/1955678396660486594) 2025-08-13T17:10Z [----] followers, [--] engagements "I avoided reading this for a while because I read about AI for many hours a week already and who needs another NYT Op Ed on it. But I succumbed after enough recommendations and I have to say I'm glad I didit's a great read. https://www.nytimes.com/2025/12/03/magazine/chatbot-writing-style.html https://www.nytimes.com/2025/12/03/magazine/chatbot-writing-style.html" [X Link](https://x.com/drmtown/status/2001132584898846955) 2025-12-17T03:29Z [----] followers, [---] engagements "I have a new post out in Lawfare today about continual learning the goal of many model developers to build AI tools that can learn from their users. That technology could have many uses but also will challenge existing ways we are trying to regulate AI" [X Link](https://x.com/drmtown/status/2001777183107461534) 2025-12-18T22:10Z [----] followers, [---] engagements "For folks who are new to the idea of continual learning I would recommend this post by @dwarkesh_sp and this post by @binarybits both of which highlight how the inability to learn puts major limits on current AI models. https://www.understandingai.org/p/context-rot-the-emerging-challenge https://www.dwarkesh.com/p/timelines-june-2025 https://www.understandingai.org/p/context-rot-the-emerging-challenge https://www.dwarkesh.com/p/timelines-june-2025" [X Link](https://x.com/drmtown/status/2001777186165133786) 2025-12-18T22:10Z [----] followers, [----] engagements "And here's a link to the Lawfare post itself https://www.lawfaremedia.org/article/when-ai-models-can-continually-learn--will-our-regulations-be-able-to-keep-up https://www.lawfaremedia.org/article/when-ai-models-can-continually-learn--will-our-regulations-be-able-to-keep-up" [X Link](https://x.com/drmtown/status/2001777187842871672) 2025-12-18T22:10Z [----] followers, [---] engagements "This by @vauhinivara is a good (and dark) chaser to the essay below: https://www.newyorker.com/culture/the-weekend-essay/what-if-readers-like-ai-generated-fiction I avoided reading this for a while because I read about AI for many hours a week already and who needs another NYT Op Ed on it. But I succumbed after enough recommendations and I have to say I'm glad I didit's a great read. https://t.co/DumBhfSbMf https://www.newyorker.com/culture/the-weekend-essay/what-if-readers-like-ai-generated-fiction I avoided reading this for a while because I read about AI for many hours a week already and" [X Link](https://x.com/drmtown/status/2003500050677473608) 2025-12-23T16:16Z [----] followers, [---] engagements "@vauhinivara (Actually "chaser" is the wrong word here. You might feel worse after reading it. But it's good to have all the different sides of the "AI writes terribly / AI writes in ways that people prefer" discussion in mind.)" [X Link](https://x.com/drmtown/status/2003500922375434727) 2025-12-23T16:20Z [----] followers, [--] engagements "Interestingly I think the information environment around capabilities is bad in law in basically the opposite direction than it seems in math. There are lots of posts/news stories about hallucinated cases while lawyers are not really incentivized to post about successes. It's good for academics to publicly experiment with new AI tools but important to report both successes and failures when doing so. Audience capture incentivizes only doing one or the other which is part of the reason the information environment around capabilities is so bad. It's good for academics to publicly experiment" [X Link](https://x.com/drmtown/status/2007182530605396025) 2026-01-02T20:09Z [----] followers, [---] engagements "RT @MarkZJia: Maduro will likely claim head-of-state immunity when he is prosecuted in SDNY. Rubio is previewing the government response w" [X Link](https://x.com/anyuser/status/2007529358802727259) 2026-01-03T19:07Z [----] followers, [--] engagements "First big civ pro case of the term is outBerk v. Choy. What's more fun than the Erie Doctrine" [X Link](https://x.com/drmtown/status/2013635676877517205) 2026-01-20T15:32Z [----] followers, [---] engagements "(The answer is: most things.) https://www.supremecourt.gov/opinions/25pdf/24-440_1b82.pdf https://www.supremecourt.gov/opinions/25pdf/24-440_1b82.pdf" [X Link](https://x.com/drmtown/status/2013635678253224391) 2026-01-20T15:32Z [----] followers, [--] engagements "Happy to see a cameo here from one of my favorite tests in all of the law: whether a procedural rule is really a procedural rule depends on whether it "really regulates procedure." First big civ pro case of the term is outBerk v. Choy. What's more fun than the Erie Doctrine https://t.co/19JE8H9dk4 First big civ pro case of the term is outBerk v. Choy. What's more fun than the Erie Doctrine https://t.co/19JE8H9dk4" [X Link](https://x.com/drmtown/status/2013636332493414659) 2026-01-20T15:34Z [----] followers, [---] engagements "RT @peterwildeford: AI makes errors. So people used AI to track how many errors AI makes. .But that also had errors. So people used A" [X Link](https://x.com/anyuser/status/2015029305865056667) 2026-01-24T11:49Z [----] followers, [--] engagements Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@drmtown Danny Wilf-Townsend
Danny Wilf-Townsend posts on X about ai, at least, thomson reuters, generative the most. They currently have [-----] followers and [--] posts still getting attention that total [---] engagements in the last [--] hours.
Engagements: [---] #
- [--] Week [---------] +16,916%
- [--] Month [---------] -93%
- [--] Months [----------] +2,283,463%
- [--] Year [----------] +94,754%
Mentions: [--] #
Followers: [-----] #
- [--] Month [-----] +0.98%
- [--] Months [-----] +0.26%
- [--] Year [-----] +0.26%
CreatorRank: [---------] #
Social Influence
Social category influence technology brands 8.33% stocks 5.56% travel destinations 5.56% countries 2.78%
Social topic influence ai 30.56%, at least 8.33%, thomson reuters #45, generative #610, law 5.56%, bloomberg #121, open ai 5.56%, georgetown 5.56%, op 5.56%, glad 5.56%
Top accounts mentioned or mentioned by @profarbel @virginialawrev @stanfordhai @adamunikowsky @vauhinivara @thezvi @garymarcus @ft @brianwolfman @dwarkeshsp @binarybits @markzjia @peterwildeford
Top assets mentioned Thomson Reuters Corporation (TRI)
Top Social Posts
Top posts by engagements in the last [--] hours
"@TheZvi Am I the only one who finds that it hallucinates like a sailor (Or whatever the right metaphor is). I still have plenty of uses for it but in my field (law) it feels like it makes it harder to convince the many AI skeptics when much-touted models make things up left and right"
X Link 2026-02-10T02:50Z [----] followers, [---] engagements
"A periodic update about the frequency and intensity of AI use in legal practice: Thomson Reuters reports that 55% of generative AI users at law firms use it at least daily with 30% multiple times a day:"
X Link 2026-02-09T20:38Z [----] followers, [--] engagements
"Of course that doesn't tell you what proportion of lawyers are "current users." Last spring Bloomberg found that a majority of lawyers "have used" generative AI for work which is at least something but isn't quite the same as "currently use.""
X Link 2026-02-09T20:38Z [----] followers, [--] engagements
"Thomson Reuters link: Bloomberg link: https://aboutblaw.com/bjbL https://www.thomsonreuters.com/content/dam/ewp-m/documents/thomsonreuters/en/pdf/reports/2026-ai-in-professional-services-report.pdf https://aboutblaw.com/bjbL https://www.thomsonreuters.com/content/dam/ewp-m/documents/thomsonreuters/en/pdf/reports/2026-ai-in-professional-services-report.pdf"
X Link 2026-02-09T20:38Z [----] followers, [--] engagements
"This is an important issue and certainly a development worth following but I would hesitate to try and learn lessons or advice for clients until there is at least a transcript or a written ruling with a sense of Judge Rakoff's reasoning. Not much to go on yet. Your AI conversations aren't privileged. Yesterday Judge Jed Rakoff ruled that [--] documents a defendant generated using an AI tool and later shared with his defense attorneys are not protected by attorney-client privilege or work product doctrine. The logic is simple: an AI Your AI conversations aren't privileged. Yesterday Judge Jed"
X Link 2026-02-12T14:39Z [----] followers, [---] engagements
"Looking forward to the philosophers of the future: can Claude create an eval so difficult that even He can't saturate it Soon enough all evals will be saturated we won't be able to make better ones and the only thing left will be vibes and "guardrails" Soon enough all evals will be saturated we won't be able to make better ones and the only thing left will be vibes and "guardrails""
X Link 2026-02-12T20:40Z [----] followers, [---] engagements
"Another important dynamic here is that even a very small proportion of lawyers misusing LLMs will result in a steady stream of stories about hallucinated cases. But it won't tell you anything about how whether / how LLMs are useful for lawyers who are acting responsibly. @GaryMarcus @FT As a lawyer who uses LLMs every day at work I feel qualified to respond. First hallucinations are no longer a problem. Consistent with the prediction you quoted from [----] GPT-5.x almost never hallucinates. And overall the percentage of inaccurate responses I get from GPT-5.2 @GaryMarcus @FT As a lawyer who"
X Link 2026-02-13T16:15Z [----] followers, [---] engagements
"I haven't spent much time with the new o3 model from OpenAI but it is the first model to get all of the questions right on the informal testing of legal questions that I've been doing for a while when new models come out"
X Link 2025-04-17T15:25Z [----] followers, [----] engagements
"The recent discovery order in Huskey v. State Farm highlights what is likely to be an increasingly common issue in "AI" cases: what specific decision-making tools are we talking about and how are those integrated into all of the other decision processes that a company has"
X Link 2024-02-26T20:11Z [----] followers, [---] engagements
"Excited to say that my article Deterring Unenforceable Terms will be coming out in the @VirginiaLawRev. It builds on the great empirical work on unenforceable terms in recent years focusing on the normative case and policy details for affirmative penalties for contract drafters:"
X Link 2024-03-07T13:54Z [----] followers, [----] engagements
"@brian_wolfman @VirginiaLawRev Thank you It's less ambitious than saying contract should be full of tort dutiesit's arguing for affirmative penalties for mass drafters which could take a few formsstatutory damages a judge-made cause of action possibly ethics sanctions for lawyer drafters etc"
X Link 2024-03-07T14:11Z [----] followers, [---] engagements
"A new study from @StanfordHAI evaluates legal research AI tools. As part of Georgetown Law's AI committee this year I've had lots of conversations with academics and practitioners about these tools and this paper highlights two important points: https://hai.stanford.edu/news/ai-trial-legal-models-hallucinate-1-out-6-queries https://hai.stanford.edu/news/ai-trial-legal-models-hallucinate-1-out-6-queries"
X Link 2024-05-29T13:38Z [----] followers, [---] engagements
"(Things can get more nuanced than just "affirm/reverse" and Claude is much better at capturing that nuance so it is actually much better than an auto-reverse program. But the baseline point is still important.) 4/"
X Link 2024-07-03T17:44Z [----] followers, [---] engagements
"But more important is the question of qualityis Claude as good as a SCOTUS clerk Here I think a ton of Claude's answers' quality comes from the briefs: SC briefs are typically excellent clear and persuasive with material for Claude to use to generate high-quality responses"
X Link 2024-07-03T17:45Z [----] followers, [---] engagements
"@AdamUnikowsky That is just pretty different from what judges and lawyers do much more often which involves not making up legal standards but determining the best answer for what existing standard applies in a given context and how. And for that a good law clerk needs to be able to look /10"
X Link 2024-07-03T17:49Z [----] followers, [---] engagements
"@AdamUnikowsky beyond the briefs and engage in legal research and reasoning without the assistance of SCOTUS lawyers clearly summarizing and writing things up. And on that front we know LLMs have a lot of problems. /11 https://x.com/drmtown/status/1795811898224517601 A new study from @StanfordHAI evaluates legal research AI tools. As part of Georgetown Law's AI committee this year I've had lots of conversations with academics and practitioners about these tools and this paper highlights two important points: https://t.co/LtDPOAAYds https://x.com/drmtown/status/1795811898224517601 A new"
X Link 2024-07-03T17:50Z [----] followers, [----] engagements
"I enjoyed reading "Seeing Like a State" and learned from it but I think everyone who likes it would also benefit from reading Paul Seabright's critical reviewa strong candidate for the best book review I've ever read: https://www.lrb.co.uk/the-paper/v21/n11/paul-seabright/the-aestheticising-vice https://www.lrb.co.uk/the-paper/v21/n11/paul-seabright/the-aestheticising-vice"
X Link 2024-07-22T14:18Z [----] followers, 32.7K engagements
"Model deletion aka algorithmic disgorgement is a remedy the FTC developed in recent years that has received a fair amount of praise with other litigants now seeking it too in AI / privacy / IP contexts. But it has a flaw: the potential to be grossly disproportionate (🧵)"
X Link 2024-08-26T13:45Z [----] followers, [---] engagements
"Where there's been a legal violation involving data model deletion requires the destruction not only of the data but also of tools trained on that data. Here is the NYT requesting it in its lawsuit v. OpenAI and the FTC requiring it in a facial recognition case (Rite Aid) 2/"
X Link 2024-08-26T13:46Z [----] followers, [---] engagements
"Charming to find out that the most popular 20th C. poet in Japan was also the translator of the (extremely popular in Japan) Peanuts comic strip. Theres no perfect analogue but imagine if Robert Frost had translated all of the Asterix or Tintin comics https://www.nytimes.com/2024/12/06/books/shuntaro-tanikawa-dead.htmlsmid=nytcore-ios-share&referringSource=articleShare https://www.nytimes.com/2024/12/06/books/shuntaro-tanikawa-dead.htmlsmid=nytcore-ios-share&referringSource=articleShare"
X Link 2024-12-12T11:04Z [----] followers, [--] engagements
"@ProfArbel This is plausible but I'm not sure it is my initial takeaway. It seems like another read might just be about market segmentation: o3 was way better than 4o but a very surprising (to me at least) proportion of people never clicked through to it and just defaulted to 4o. So "
X Link 2025-08-13T16:41Z [----] followers, [--] engagements
"@ProfArbel .those people who apparently are a supermajority of ChatGPT users are likely to experience [--] as a significant step up because they didn't (or rarely) encounter reasoning models in their previous uses"
X Link 2025-08-13T16:42Z [----] followers, [--] engagements
"@ProfArbel So I have been wondering if the main updating here to my thinking should maybe be "OpenAI cares more about the median consumer" than I had thought (vs. caring about consumers at the frontier of capabilities). As opposed to updating on e.g. talent diaspora"
X Link 2025-08-13T16:46Z [----] followers, [--] engagements
"@ProfArbel Yes definitely agree that these aren't mutually exclusive. Re the mad scientist frontier lab one thing I'll be looking out for is what ends up happening with the more intelligent model they claim to have internallywhen/how is that released how does it end up doing etc"
X Link 2025-08-13T17:10Z [----] followers, [--] engagements
"I avoided reading this for a while because I read about AI for many hours a week already and who needs another NYT Op Ed on it. But I succumbed after enough recommendations and I have to say I'm glad I didit's a great read. https://www.nytimes.com/2025/12/03/magazine/chatbot-writing-style.html https://www.nytimes.com/2025/12/03/magazine/chatbot-writing-style.html"
X Link 2025-12-17T03:29Z [----] followers, [---] engagements
"I have a new post out in Lawfare today about continual learning the goal of many model developers to build AI tools that can learn from their users. That technology could have many uses but also will challenge existing ways we are trying to regulate AI"
X Link 2025-12-18T22:10Z [----] followers, [---] engagements
"For folks who are new to the idea of continual learning I would recommend this post by @dwarkesh_sp and this post by @binarybits both of which highlight how the inability to learn puts major limits on current AI models. https://www.understandingai.org/p/context-rot-the-emerging-challenge https://www.dwarkesh.com/p/timelines-june-2025 https://www.understandingai.org/p/context-rot-the-emerging-challenge https://www.dwarkesh.com/p/timelines-june-2025"
X Link 2025-12-18T22:10Z [----] followers, [----] engagements
"And here's a link to the Lawfare post itself https://www.lawfaremedia.org/article/when-ai-models-can-continually-learn--will-our-regulations-be-able-to-keep-up https://www.lawfaremedia.org/article/when-ai-models-can-continually-learn--will-our-regulations-be-able-to-keep-up"
X Link 2025-12-18T22:10Z [----] followers, [---] engagements
"This by @vauhinivara is a good (and dark) chaser to the essay below: https://www.newyorker.com/culture/the-weekend-essay/what-if-readers-like-ai-generated-fiction I avoided reading this for a while because I read about AI for many hours a week already and who needs another NYT Op Ed on it. But I succumbed after enough recommendations and I have to say I'm glad I didit's a great read. https://t.co/DumBhfSbMf https://www.newyorker.com/culture/the-weekend-essay/what-if-readers-like-ai-generated-fiction I avoided reading this for a while because I read about AI for many hours a week already and"
X Link 2025-12-23T16:16Z [----] followers, [---] engagements
"@vauhinivara (Actually "chaser" is the wrong word here. You might feel worse after reading it. But it's good to have all the different sides of the "AI writes terribly / AI writes in ways that people prefer" discussion in mind.)"
X Link 2025-12-23T16:20Z [----] followers, [--] engagements
"Interestingly I think the information environment around capabilities is bad in law in basically the opposite direction than it seems in math. There are lots of posts/news stories about hallucinated cases while lawyers are not really incentivized to post about successes. It's good for academics to publicly experiment with new AI tools but important to report both successes and failures when doing so. Audience capture incentivizes only doing one or the other which is part of the reason the information environment around capabilities is so bad. It's good for academics to publicly experiment"
X Link 2026-01-02T20:09Z [----] followers, [---] engagements
"RT @MarkZJia: Maduro will likely claim head-of-state immunity when he is prosecuted in SDNY. Rubio is previewing the government response w"
X Link 2026-01-03T19:07Z [----] followers, [--] engagements
"First big civ pro case of the term is outBerk v. Choy. What's more fun than the Erie Doctrine"
X Link 2026-01-20T15:32Z [----] followers, [---] engagements
"(The answer is: most things.) https://www.supremecourt.gov/opinions/25pdf/24-440_1b82.pdf https://www.supremecourt.gov/opinions/25pdf/24-440_1b82.pdf"
X Link 2026-01-20T15:32Z [----] followers, [--] engagements
"Happy to see a cameo here from one of my favorite tests in all of the law: whether a procedural rule is really a procedural rule depends on whether it "really regulates procedure." First big civ pro case of the term is outBerk v. Choy. What's more fun than the Erie Doctrine https://t.co/19JE8H9dk4 First big civ pro case of the term is outBerk v. Choy. What's more fun than the Erie Doctrine https://t.co/19JE8H9dk4"
X Link 2026-01-20T15:34Z [----] followers, [---] engagements
"RT @peterwildeford: AI makes errors. So people used AI to track how many errors AI makes. .But that also had errors. So people used A"
X Link 2026-01-24T11:49Z [----] followers, [--] engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing