Pete

@spacerog @HockeyInJune @vyrus001 @Gillis57 @HushCon The math doesn't work for researchers. Here is the most glaring example I've found where disclosure increases risk. (Btw, the collision arg holds no weight either.)

@ruijianggao This is very interesting. I wonder if #2 will end up with the same issue as p-hacking. Thoughts?

Takeaway for social scientists the AI era: 1️⃣ Never trust a single AI estimate. 2️⃣ Use Multiverse Analysis: run many agents to see the full range of plausible results. 3️⃣ AI NSE may serve as a "lower bound" for human NSE.

What happens when you invite 150 AI economists (Claude Code) to a research conference, give them the exact same data, and ask them to test the same hypotheses? We did just that. The results reveal a new phenomenon: Nonstandard Errors in AI Agents. 🧵👇

@jessesingal @eigenrobot

@jessesingal @eigenrobot

Isn't there a fairly famous passage in Rise and Fall of the Third Reich where he talks about true-believer Nazi and Soviet types getting along really well? I don't (think I) mean the section about Ribbentrop in Moscow...

@royalhansen I wonder if pragmatic interp will affect this interest at all. Any thoughts?

Another great piece from my friend Allen - this observability gap also keeps us from the fully self-defending apps we need allen.hutchison.org/2026/02/17/the…

@alexolegimas @Luigi1549898 @R_Thaler Ah, yes, I got the earlier version of "Winners Curse." Will have to get the update. Of course, another model of the stock market might be "turtle all the way down..." 😀

@SpireSec @Luigi1549898 @R_Thaler Yup @R_Thaler and I talk about this as a model for the stock market in our book.

I’m not a lawyer so I won't interpret contracts for you. But I am an economist, so here's my take on the economics of what happened yesterday (TL;DR: Jesse is right). One of the main drivers of economic activity in developed countries is a favorable regulatory environment and *certainty* about that environment. Any investment is risky, and regulatory/institutional uncertainty keeps companies from investing, growing, and building in these types of environments. Companies don't want to build a new plant if they think the govt will seize it or put extra regs on its products. As an example, take the tariff bonanza of the past year +. It's one thing to enact a 15% across the board tariff and stick with it, and entirely another to draw up completely random numbers, change them constantly (e.g., because you don't like a commercial), all under the spectre of them being illegal in the first place (they were). The outcome was exactly as predicted: consumers paid most of the cost, manufacturing is shedding workers and pulling back on construction (after 4 years of increases). The economy does not like uncertainty. The one bright shining star of the economy was AI. The US had two of the frontier labs and the most of the major tech companies. AI had the most favorable regulatory environment and was booming; investment was going through the roof. And then the secretary of defense decided to nuke the business model of one of those companies because of too many stupid reasons to list here. It would be perfectly fine to cancel their contract, but designating Anthropic a *supply chain risk* --which has so far been reserved for foreign adversaries -- cuts off much of their business. But so what? Anthropic is important, but it's just one company. The problem is that this overturns the regulatory certainty that has made the US such a beacon for AI. AI companies, which have a lot of options, will now think twice unless a capricious member of the admin wants to designate an important part of their supply chain risk. Maybe this will all blow over, maybe there will be a walk back. But trust is hard to build, but very easy to break---especially when it comes to economics and risk. I hope I'm wrong.

@MikeBerardino Jadarian Price is a great teammate...

My latest on #NotreDame: "Jeremiyah is a freak. Getting to block for him is a blessing. #NDFootball #NFLCombine southbendtribune.com/story/sports/c…

@alexolegimas @Luigi1549898 Reminds me of the @R_Thaler "riddle" to pick a number between 1-100 that is 2/3 of the average of the population responses to this question.

@Luigi1549898 Yeah it’s hard for me to interpret the stock market as a signal of anything these days, it’s levels upon levels upon levels of reasoning.

@NateSilver538 Idk, maybe do a search on "remembered by historians" in the past and see what the likelihood is?

I don't think this is a particularly bold assertion?

@growing_daniel The pendulum for the Turing test has swung in the other direction. Many don't realize that AI "beating" humans at being human - being identified at >50% - is also a failure (and interesting indicator of human bias). Now we need AI to administer and evaluate the test.

People act like the Turing test is defeated and it’s really not. We all recognize AI writing fairly easily.

@mattjay @anton_chuvakin Do me a favor and keep that down for a few days 😉.

The market is an idiot. Crowdstrike doesn’t even find or fix vulns. People just think cyber is cyber. I’m not a financial advisor but Claude’s announcement impacts Crowdstrike 0%

A Republic, if you can keep it marginalrevolution.com/marginalrevolu…

@EPoe187 @NoahCarl90 This is why Badler and Grinder started nlp.

"The idea that understanding the cause of one’s suffering is the key to curing it is dubious. Getting terminated from a job might cause misery but ruminating on one’s termination is unlikely to dissipate one’s depression—and may exacerbate it." aporiamagazine.com/p/the-psychoth…

@gmeta @PostOpinions @tylercowen @asymmetricinfo @grok Only Boomers know what "boomer" usually refers to 🤣.

@PostOpinions @tylercowen @asymmetricinfo @grok is WAPO saying that somehow a doomer is the opposite of a boomer?

Economist and professor @TylerCowen chats with Megan McArdle (@asymmetricinfo) about how AI could transform talent, human capital and competition: wapo.st/4aibIz2

@beaverd @NickSzabo4 Go six levels to get to Kevin Bacon.

2.1 million relationships, 1.5 million nodes from the Epstein files are now in a 3d graph to explore connections and export data. nodes are up to 3 layers deep, 13,800 orgs linked, 38,000 people somaliscan.com/investigation/…

don't worry guys, they're just stochastic parrots

@emollick Do you expect LLM success to continue or some other form of AI? Seems like very prominent people - Li, LeCun - looking for what's next combined with the paper(s) that claim LLM math doesn't work in the face of complexity might be good indicators that brute force is topping out.

The many eulogies for AI capability growth after the release of GPT-5 seem especially short-sighted right now. Letting people nervous about AI feel they can safely ignore AI development because it was pure hype that would never have any real impact is not a good thing for anyone

@chr1sa @paulg Don't you at least need to add in YouTube and TikTok?

One of my hobbies since writing The Long Tail is tracking the decline of blockbuster culture in the US. It's striking how film streaming is still not even close to compensating for the decline of DVDs and broadcast

@JamesSurowiecki I'm worried about Munchausen Trump. He's been creating systemic problems and now may try to "solve" them with superficial measures to address them in order to be seen as a hero. Some small subset of people benefit but the entire country/world is worse off.

There are two obvious patterns at work here. First, MAGA candidates other than Trump are deeply unpopular with voters anywhere other than deep-red districts, and sometimes even in those districts. Second, Trump's shtick has worked reasonably well during campaigns, because voters find him entertaining and a welcome alternative to the status quo. But once he takes office, and governs as the narcissistic chaos agent that he is, voters quickly come to dislike him.

Trump-endorsed Republican candidate and MAGA activist loses by 14 points in a district Trump won by 17 points in 2024. I'm sure Trump will erase this loss from his memory when he talks about how candidates he endorses always win. nytimes.com/2026/02/01/us/…

@greg2126 Well, Williams lost tonight, too.

That play reminded me Lincoln Riley went 7-5 with him as his QB.

@pescami Idk, am I fool for being confused here? Vance or Rubio?