Hacker Newsnew | past | comments | ask | show | jobs | submit | brokensegue's commentslogin

jython has been basically unmaintained for quite some time

Well, they never made the jump to Python 3. But shipping 2.7 interpreters in 2024 was quite an achievement on its own. So their users already know this pain. And from my experience in academia, python 2.7 and java 8 will probably be used for another 20 years before the last machine running that stuff burns out.

Their variable cost is (basically) the number of tokens. They increased that. I don't get how that saves them money


You can freeze and concentrate a substance without chemically altering it


none of that is censorship


I think b/w film has a different grain than color. It isn't identical to grayscale color


many people do not think the science against them are credible. foods containing those things are staples of many diets.


i've tried it a few times and it's never really helped as much as i expected. though i know they've released a couple times since I last tried it.


yeah what I'm trying to get across here is that: Dspy does not solve an immediate problem, which is why many feel this way and consequently why it doesn't have great adoption!

But on the other hand, I think people unintentionally end up re-implementing a lot of Dspy.


genie is out of the bottle on this one


they used to claim that being Swiss based protected them from warrants like this


Source? We need the exact claim here, because there's a fine line between "we're in switzerland, so warrants aren't a thing!" (outright false) and "we're in switzerland, which have better privacy laws than other countries!" (debatable).


I’m not the person who made the claim, but a basic web search led me to this page on their blog:

https://proton.me/blog/data-privacy-abortion

Quote (emphasis theirs, in bold):

> Switzerland is a fundamentally different environment. Two of the things Switzerland is most famous for are also highly conducive to data protection: privacy and neutrality.

> When a law enforcement agency in the US requests user data from a Swiss company, it is illegal for that company to provide the data. At Proton, we reject all data requests from foreign agencies.

> Proton and other Swiss companies will only hand over user data when ordered to do so by a Swiss authority. And even then, Proton’s general policy is to challenge data requests whenever possible and only comply after all legal remedies have been exhausted.

So maybe your parent poster is confused? They do claim that being Swiss protects them from requests from foreign entities, but not Swiss entities. Which is what happened here, the Swiss authorities asked Proton for the data, then they handed it to the FBI.

Has Proton challenged the data and “only complied after all legal remedies have been exhausted”, though? That’s another question.


I wonder if the FBI knew it was going to be a pain in the ass asking for actual account access from the Swiss so they asked for financial records instead. Terrorism charges look pretty serious (regardless of how legitimate they are) so I'm sure that's what pushed the Swiss and Proton to comply.


But the Swiss have the notion of a warrant, no? So if a Swiss judge or official issues a proper warrant, then a Swiss company or citizen is obliged to comply with it.


Swiss police requested the data and handed it to the FBI.


Read the wording, they aren't complying with US warrants, they are complying with Swiss-issued warrants. US LE/Intel agencies figured out this loophole some 20 years ago.


Citation needed. They really said they were above Swiss law?


"classical ML" models typically have a more narrow range of applicability. in my mind the value of ollama is that you can easily download and swap-out different models with the same API. many of the models will be roughly interchangeable with tradeoffs you can compute.

if you're working on a fraud problem an open-source fraud model will probably be useless (if it even could exist). and if you own the entire training to inference pipeline i'm not sure what this offers? i guess you can easily swap the backends? maybe for ensembling?


> if you own the entire training to inference pipeline i'm not sure what this offers

336x faster than Python, and swapping backends in a production environment can be far from trivial


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: