HN user
Computer science academic.
I find it kind of funny that the last time a partial member list leaked (a few years ago) it was because they tried to invite Andrew Gelman and he mocked it on his blog: https://statmodeling.stat.columbia.edu/2022/02/16/hey-i-got-...
I guess he didn't get invited again.
Any of the direct indexing providers will let you blacklist individual stocks from the index. The intended use is to exclude stocks you hold elsewhere (or receive as stock grants) to avoid causing wash sales, but it can also be easily used to make a custom "S&P 499".
The article does discuss union collective bargaining agreements and labor cost structure in several sections.
Labor costs might seem variable, but they’re actually not: pilot, flight attendant, and mechanic compensation in the United States is governed by the Railway Labor Act of 1926 (which was extended to airlines in 1936), which stipulates that collective bargaining agreements don’t actually expire but rather remain in force until they’re replaced. So even your wage bill is more or less fixed over multi-year horizons.
Chapter 11 bankruptcy protection—which allows a company to continue operating while it restructures its debts under court supervision—is practically the only mechanism by which an airline can renegotiate its rigid cost structure, from aircraft leases to collective bargaining agreements. Oftentimes this renegotiation takes on a rather predatory character. When United Airlines filed for bankruptcy in 2002 in the aftermath of the September 11th attacks, it terminated its pension plan...
So Chapter 11 is a relief valve for airlines struggling under the weight of their fixed costs; but it doesn’t really do much to help the system as a whole. For airlines, bankruptcy rarely culminates with liquidation; airlines that emerge from bankruptcy proceedings, having voided pension obligations and rejected aircraft leases, can operate at a fundamentally lower cost basis than their competitors. So bankruptcy doesn’t really restore the industry to a competitive equilibrium that can cover the cost of capital: it resets the floor at a lower level, from which a new round of ruinous competition can begin.
Hard to square that with how good open-weights models are getting? I'm doing stuff with Qwen3.5-4b that required a frontier hosted model less than a year ago.
It's not just an inner city issue. Rural areas in the U.S. are bad too! Parts of the rural south and rural midwest have homicide rates that are completely off the charts for a first-world country (Holmes County, MS; Scioto County, OH; McDowell County, WV).
The key part to me:
Congress wrote to the Secretary of Defense about the extent of AI use in the strikes. The New Yorker asked whether Claude could be trusted to obey orders in combat, whether it might resort to blackmail as a self-preservation strategy, and whether the Pentagon’s chief concern should be that the chatbot had a personality. Almost none of this had any relationship to reality. The targeting for Operation Epic Fury ran on a system called Maven. Nobody was arguing about Maven.
Thank you for the link! Now that is the article I wanted on this topic.
The article didn't answer my main question, which is how the economics work. How does it add up to have high-touch home delivery of $5 yogurt packages?
Largely in-person gigs in the SF Bay Area, yeah. One reason the rents are up so much.
Economy-wide productivity can be measured reasonably well, although there are a few different measures [1]. The big question I guess is whether AI will make a measurable impact there. Historically tech has had less impact than people thought it would, as noted in Robert Solow's classic quip that "You can see the computer age everywhere but in the productivity statistics". [2]
[1] https://www.oecd.org/en/topics/sub-issues/measuring-producti...
those new self-checkout shopping carts
I'm going to miss those. Two nice things about them compared to a normal self-checkout: 1) you see things ring up as you shop instead of at the end, which is nice in case of errors or unexpected prices, 2) you can shop directly into a reusable bag or backpack instead of repacking everything at the end.
Both sides here are $100B+ entities! But neither one is acting like it. The plaintiff is shitposting on Twitter, and the respondent is counter-shitposting on their blog (and also Twitter).
There are many unresolved gray areas around what exactly the 4th amendment permits in the way of what United States v. Knotts called "dragnet-type law enforcement practices". Knotts suggested they might not be permitted, even if they were made up of permissible individual parts, but didn't elaborate. More recent case law has held, for example, that cell phone companies turning over large quantities of records is a 4th amendment search requiring a warrant, even if they do it voluntarily (https://en.wikipedia.org/wiki/Carpenter_v._United_States). Most other types of dragnets haven't been litigated enough to have solid caselaw on their boundaries afaik.
I don't know if it's likely a court will do anything about this particular program, but from what I've read I don't think 4th amendment scholars think this area is at all settled.
I've got more less tips than the Bible's got Psalms
But there are (at least) 150 Psalms! You're going to need more less tips to match that.
That's good to know, thanks. In my case nearly 100% of my code ends up public on GitHub, so I assume everyone's code models are training on it anyway. But would be worth considering if I had proprietary codebases.
As a third option, I've found I can do a few hours a day on the $20/mo Google plan. I don't think Gemini is quite as good as Claude for my uses, but it's good enough and you get a lot of tokens for your $20. Make sure to enable the Gemini 3 preview in gemini-cli though (not enabled by default).
The US is more like Latin America here, where top 25% does not mean you are immune to the problems of the other 75%. Your average resident of an affluent western European country does not come into daily contact with a poor resident of western Russia (except maybe through some recent drones...?). But in the US, it's hard to escape the poverty and violence unless you're closer to the top 1%. I am probably around top 20%, in an affluent part of an affluent coastal US city (but not Manhattan affluent). Still, I don't walk outside after dark because of the high crime rates. The crime rates are not as high as in the bad parts of the same city, but people are mugged on a daily basis, and even murders are uncomfortably frequent. I didn't have this problem when I lived in several different places in western Europe, which had lower incomes, but also much lower levels of violent crime. I will admit that American pay is better, but not sure about the quality of life.
Interesting to see Thiel putting forth Eliezer Yudkowsky as the possible antichrist. Hasn't Thiel himself given Yud/MIRI like $2 million?
I've found myself getting less interested in sports at all because of how pervasive sports betting has gotten. The announcers are always talking about odds and shilling gambling company sponsors, which is annoying and makes me not want to watch the games.
Inkjet printers that can do 11x17" are usually marketed under a "pro" or "business" line, but not necessarily expensive. Here's a $279 model: https://www.bestbuy.com/product/epson-workforce-pro-wf-7310-...
There are local LLM coding models that ship with XCode now too.
There have in fact been some results on quantum speedups for machine learning: https://www.quantamagazine.org/ai-gets-a-quantum-computing-s...
I would expect this to become relevant later than crypto, though, because you need larger data sizes for things to get interesting.
It’s definitely common in Denmark, enough that it’s a perennial national debate. Maersk is a huge employer that officially made their corporate language English something like 20 years ago, which spawned discussions about whether you should need to speak a foreign language to get a job in your own country. In practice the answer is yes, for some sectors.
I worked for years in an English-language work environment in Denmark (I am not Danish), and learned maybe a handful of phrases of spoken Danish the entire time. I was expected to be able to read the occasional email in Danish, but 1) written Danish is not hard in comparison, and 2) even years ago Google Translate was good enough.
It would have been nice from a social perspective to have known more spoken Danish, but my employer didn’t really care, and it isn’t easy to learn if you don’t have strong local connections. Danes will just immediately switch to English by default, and even if you ask them to continue in Danish, you need a decent level of Danish pronunciation to make yourself understood, which is not trivial to get to.
For this book specifically, I think the reason is just that Lean is the theorem prover he already knows and uses in his own work [1].
Now as to why he picked Lean a few years ago as his go-to theorem prover, I do think that would be an interesting backstory. As far as I know he hasn't written anything explicit about it.
[1] E.g., https://terrytao.wordpress.com/2023/11/18/formalizing-the-pr...
I agree, especially because in practice the vast majority of AI-generated videos uploaded to YouTube are going to be from one of about 3 or 4 generators (Sora, Veo, etc.). May change in the future, but at the moment the detection problem is pretty well constrained.
I don't think it's been dead for 10 years. I'd place it at maybe 3? I teach at a mid-ranked university and the majority of my fresh out of college students were getting good-to-great entry level offers just a few years ago. The top 5-10% were getting well into six figure offers from FAANG or similar companies. But the entry level job market really tanked in mid 2022 when all the big tech companies did rounds of layoffs, and it's been much harder since then.
Genetic programming systems have periodically made improvements to algorithms (dating back decades). Whether LLM-powered GP, which is effectively what this is, will be a step change or an evolution of that is still an open question I think. I'm also a little wary of reading too much into the recursive self-improvement idea, because "the GP system can use GP to improve the GP system itself!" is a very old idea that just has never worked, although I realize that isn't proof that it won't eventually work.
Some related work from a different company: https://sakana.ai/ai-cuda-engineer/
And some academic papers kind of in this space: https://arxiv.org/abs/2206.08896, https://arxiv.org/abs/2302.12170, https://arxiv.org/abs/2401.07102
This might have changed after you posted your comment, but it looks like 2.5 Pro and 2.5 Flash are available in the Gemini app now, both web and mobile.
Indeed meteofrance appears to be monolingual. I would say they're a bit unusual in that though. Most other national meteorological services I know of have forecasts available in at least two languages.
E.g.:
Finnish/Swedish/English: https://www.ilmatieteenlaitos.fi/, https://sv.ilmatieteenlaitos.fi/, https://en.ilmatieteenlaitos.fi/
Italian/English: https://www.agenziaitaliameteo.it/, https://www.agenziaitaliameteo.it/en/
A dozen or so languages (full forecasts available in about half): https://www.jma.go.jp/jma/kokusai/multi.html
I believe the implication is that they're seeding the problems in the discourse for the purpose of selling their solutions. A slightly more overt version of the submarine: https://www.paulgraham.com/submarine.html