But the fact that she went on leave just as it became clear that Anthropic looked to be hitting escape velocity is suspicious, although not dispositive. When she was hired, her health issues also stopped her from coming into the office, but she still worked from home.
The Noah Smith/Nate Silver axis of opinion-politics writers are — literally — Stephen J Smith and Skip Bayless for people who think politics is sports. They're in the entertainment business.
There end up being some weird edge cases where there are some countries which have both the equivalent of .co.uk but also allow registrations directly under the two-letter country code as well. .mx is one such case where most business are, e.g., costco.com.mx, but it’s also possible to register directly under .mx as well so Toyota Mexico is toyota.mx and not toyota.com.mx (the latter is registered, and ostensibly to Toyota, but the whois and nslookup records give very different results and the website doesn’t load when I try to visit it).
This isn't so bad as .com.mx and .mx should be on the public suffix list then.
But letting arbitrary customers take arbitrary 3 level domains, and others take 2 level domains, seems like a mistake as it's not very reasonable for every 3LD customer to put the 2LD on the public suffix list, but mixing 3LD and 2LD registrations means you can't public suffix *.name.
Seems the whole idea of having both was always misguided.
The .us domain should’ve been universally useful for state and municipal governments, but most of those began registering directly under .gov, and not even in an orderly hierarchy under .st.gov
But that was simply the easiest way to market your website as a trusted government entity. And now nobody has ever heard of .us domains in active use.
.us was primarily a hierarchy structure which in practice made confusing and hard to remember domain names, whereas .gov addresses hand out single domains which are generally easy to remember.
Personally, I never saw anything confusing about city.state.us; the hierarchy was organized perfectly logically in the 3-tier jurisdictional structure that every American schoolboy knows by 3rd grade.
But your point about them being rather longer and difficult to remember stands, and the same for a .gov, which could be shorter and catchier.
However amusingly, .us opened up second-level registrations 24 years ago, which means that any qualifying entity could have their name registered directly under .us, which is obviously recognizable, and also one character shorter, than a .gov registration. However, by that time, I believe that .gov had increased in stature so that registering governmental entities under .gov carried more certainty of conveying official status than anything under .us.
Also sadly, QR Codes and URL shorteners today sort of obviate the need to directly register the shortest possible domain name. I don't know: I was always kind of fond of the .us hierarchy, and I'm just personally sad that it's fading away.
City/state/US is logical, the problem is most other hierarchies confuse people. For instance k12 subdomains for schools couldn't use that nomenclature because school districts do not map cleanly to towns. And that's before you talk about fire departments, townships, libraries, park districts, and countless other governmental bodies and districts which have overlapping boundaries of their own.
.gov certainly cares a level of exclusionary access that isn't really true of .us. Only one entity, the US federal government, can decide to hand someone a .gov address. And generally there is few signals harder to fake or impersonate than one.
It’s not something anyone else in the world seems to struggle with, where there are *.gov.uk, *.edu.au etc.
If anything the .gov, .mil and .edu being just American is confusing, as well plainly inappropriate (it feels like an American cultural imperialist thing to people from outside the US). It would have been much better if those had been retired decades ago and moved to under the .us TLD, so e.g. whatever.edu would become whatever.edu.us like every other country. Any existing domains on .gov, .mil, .edu etc. should only be allowed to exist as 301 redirects.
It's imperialist to own and control the thing you created?
If .gov had been an international TLD that was at some point available to everyone, or had been created by everyone, ok. But .gov was created as part of the work the US government did to build out the initial DNS structure. It probably wasn't even a given at the time that arpanet would be international in nature
Also, 301 redirects are not a DNS thing, that is an HTTP thing. Not sure how that would solve your problem since HTTP is intrinsically at the base of it tied to just A or AAAA records. DNS does a lot more than pointing to websites
It used to be that only Japanese corporations could register a .co.jp while anyone else anywhere could register for a .jp. So I had several .jp domains registered through Gandi.net.
The issue is that .jp registered outside of a few Japanese registrars are legally not allowed to offer Whois privacy.
Based on my small sample of schools, all of the ones that were using locality based names under ca.us have migrated elsewhere, including to 2nd level domains under .us.
The implication (given that the US is an outlier here) is that prevention is much better than cure here; that means reform of healthcare and social housing so people are less likely to wind up in that trap.
Yes you can able to see IPv6 at your browser, but actualy there is always middlewhere sofware level (NLB, ALB, NAT) to reach real backend server, which creates bottle neck.
For large scale enterprises this creates issue, even for cloudproviders, I have encountered bottle neck at some Major cloudprovider which causes snowball impact to whole cluster retry mechanisms and 4 hours outage for us and some other customers :D
if the 2 enterpise has machine to machine IPv6 connectivity, this is will not happend.
What is the error rate of human programmers? Anyone who tells you that either human or A.I. code is magically exempt from issues is selling you something.
Mmm there are humans who think they make very few mistakes. Ones who actually make few mistakes, not sure about that one. Could be a mistake that they catch themselves very quickly, but I just don’t think humans are very good at generating 100% reliable output on the first try anywhere close to most of the time.
Other post-train mechanisms have meaningful bandwidth for introducing new information to the model (on-policy distillation is more or less the other extreme). RLVR aligns the model to particular behaviors it could already (unreliably, intermittently) express, it doesn't introduce new behaviors; the behaviors were latent in the pretrain/midtrain.
This paper is less dramatic than you think it is and really just re-explains what RLVR does.
Let's stipulate that what pretraining does is train next token prediction over a gigantic corpus. You can then sample from this distribution repeatedly (cf the Large Language Monkeys paper) and count how often it passes some deterministic verifier.
What GRPO-style RLVR does is precisely this, but then reward the trajectories which passed the verifier. These distributions are _by construction_ within the accessible output space of the pretrained model; you're reweighting the distribution so that pass@k goes up, because that's (for applications like programming) very useful. RLVR is about making sampling more efficient; the only new information being added to the system is the presence of the verifier, and note that you only get a reward when the verifier passes, so there's essentially no mechanism for "teaching new facts" here.
reply