freehorse

4 ms·

61.
by freehorse 1mo ago
A 3x price difference means you can get 3 256GB mac studios which you can connect through thunderbolt and with RDMA a total of 768GB ram with compute/memory bandwidth scaling basically linearly (with a small overhead cost).
62.
by freehorse 1mo ago
Except for the prices of course, where they use “starting at 5,499$” instead of “up to 18,299$” to indicate user options.
63.
by freehorse 1mo ago
This is my impression also, there is usually explicit "only customers" where there are genuine problems (overtouristic etc) or where the boss is just weird (even though you may be still told off even without a only-customers sign
64.
by freehorse 1mo ago
It sounds more like "the consequences of this regulation are your fault because you did not protest enough".
65.
by freehorse 1mo ago
When big companies have the resources to lobby and pass regulations (EU or country wide level) it is hard to separate these two, in particular when there are long-lasting policies that consistently disadvantage smaller entrepreneurs/sm
66.
by freehorse 1mo ago
The fact that these clothes tend to be in the category of "doesn't last or made of plastic" (plus low human labour costs) is what makes them cheap enough to be possible for people to order/buy them in as large quantities
67.
by freehorse 1mo ago
"EU" is too broad of a term to describe cultural/social/political norms such as public toilets. There are places in the EU with fairly decent toilets, and others with almost none. In some you have to pay in many public t
68.
by freehorse 1mo ago
Homeless heroin addicts who inject themselves in public are not uniformly distributed. Thus while in some places you would expect some such injectable drug use in a public toilet (though I doubt most would bother to go find a toilet for tha
69.
by freehorse 1mo ago
That's against even more conventions, even though logically it is not wrong.
70.
by freehorse 1mo ago
To be fair, it does say > If your app doesn’t handle any types of files, you can rename or eliminate this menu.
71.
by freehorse 1mo ago
- Notes: note - Messages: message - Passwords: password - Reminders: reminder [...] Most of propositions are restatement of the app name or theme thereof, and thus do not seem to be very helpful (everything in a notes app has to do with &qu
72.
by freehorse 1mo ago
it works on orion
73.
by freehorse 1mo ago
With qwen 27b, setting the right reasoning effort for the specific task is important. With xhigh it has a chance at hard problems that bigger models may even fail. But for many everyday tasks, I have found that no reasoning and a system pro
74.
by freehorse 1mo ago
Previous discussion on the github issue thread mentioned: https://news.ycombinator.com/item?id=49218707 Archived page of said github thread itself: https://web.archive.org/web/20260731053721/http:&
75.
by freehorse 1mo ago
Due to that image, I immediately copy pasted it to pangram before reading the author's note. Interestingly enough it marked it all as AI generated that has "been paraphrased or rewritten". I am not sure I trust that though,
76.
by freehorse 1mo ago
They could just have no "featured image"? Quasi-irrelevant featured images where already bad before AI, AI-generated ones are just 10x worse. Anyway, it is their piece, their website, so they choose how to structure it. I just don
77.
by freehorse 1mo ago
What AI definitely is as an amplifier (in my interpretation of the article) is an amplifier of "laziness": students spending less time studying, getting by with the homework with the less possible effort, and finally achieving wor
78.
by freehorse 1mo ago
I meant, they are not gonna start from scratch by crawling websites, they are gonna rely on the search results to grab content of relevant results. I think this is what research does, in a higher quantity. In contrast, I may start with a pa
79.
by freehorse 1mo ago
LLMs are not magic, nor are they gonna crawl some dozens of websites, they work with search results they are provided with, and thus they need good search results to work with. Maybe in the US or sth it is better, but where I live finding g
80.
by freehorse 1mo ago
> Traditional search days are really numbered. There is a lot that agents can help with wrt search, but they also have a lot of blind spots. They work well many times, but then there are times where I find stuff that multiple llms missed
81.
by freehorse 1mo ago
I often try 2 different models with the same prompt in case one misses anything. But most of the times I do not see much difference. Newer models do not seem to improve search that much, so I do not think you have to stress too much about n
82.
by freehorse 1mo ago
I have had issues with an otherwise decent 2022 computer, had to arrange usb connections in specific ways to avoid brief connectivity drops. It did not happen all the time, but it did happen. But the worst problems were older ones. USB port
83.
by freehorse 1mo ago
Dongles are indeed not perceivably different than wired in principle. However, you may have more latency variability, or even unstable connectivity, depending on what adjacent usb ports are doing, due to interference [0]. I have had cases w
84.
by freehorse 1mo ago
I also believed that, but seeing qwen 27b overengineering solutions in a bit too familiar way in the article, I started doubting that.
85.
by freehorse 1mo ago
Good thing that in our universe L-glucose seems to act as a laxative, so there is less of an incentive to mass-produce that for dieting purposes.
86.
by freehorse 1mo ago
In my google search, it gave me directly a translation box where it was translated as "personal scale" (which I think is also used).
87.
by freehorse 1mo ago
I assume it is due to machine translation? I do not see mentions of "failure" at all, which would be expected to exist if it was to increase variety. I think (heard) some used to translate an article back and forth a few times to
88.
by freehorse 1mo ago
On the other hand, a motor you can strap can also be unstrapped, and then you can use the bike as a normal bike. When you have an electric bike, it is usually too heavy to work comfortable without battery (ime). I am not sure if this is eno
89.
by freehorse 1mo ago
I have found that q8 kv quantisation is really bad in anything requiring some kind of literal retrieval even at relatively shorter contexts (5-10k). Eg it would mess up names, titles of papers etc that were deeper in the context.
90.
by freehorse 1mo ago
If you do not want to tweak parameters you can just run llama-server -m model.gguf That's it

More ›