desterothx

5 ms·

31.
by desterothx 1mo ago
You state it as if its only the quality of the hypothesis that matters, but you are ignoring an important part of it, timing. During the 08 financial crisis Burry had a hypothesis that was correct, however he almost went bankrupt still beca
32.
by desterothx 1mo ago
Yes, and it still wouldnt help you since they are trying to fetch everything for each of the 921 forks, as stated in the article (same commits need to be cached again on a forked branch)
33.
by desterothx 1mo ago
If everyone had devices that could do some compute worth paying for, people would be doing the work all of the time. The problem is actually useful POW is a lot more expensive compute-wise, making it non viable for normal users
34.
by desterothx 1mo ago
The use of LLM agents was allowed relatively recently, you still have 20 years of guaranteed no llm commits in the history. As for personal LLMs being the cause of the traffic, I think you are overestimating the number of users of LLMs gene
35.
by desterothx 1mo ago
It's not a hard problem, the hard part is doing it while keeping a similar level of privacy/anonimity
36.
by desterothx 1mo ago
im rather fond of a blood tester that gives you unique drug cocktails as a fingerprint
37.
by desterothx 1mo ago
I vote for a finger pricker that tests the blood composition everytime I want to look up the nearest open grocery store
38.
by desterothx 1mo ago
The trouble is still differentiating between real users and agents. Sure this could be useful (forcing agents to spend a certain amount of compute towards something actually useful), however real users that get stuck with it would be stuck
39.
by desterothx 1mo ago
I mean generally what you said is true. But who is actually going to look at the linux kernel from their phone (probably just the maintainers in extreme circumstances)
40.
by desterothx 1mo ago
The troubling part for me is that a system like this could also just as easily be pawned onto the humans (I realize it technically already is with the subscription model internet, but there are still plenty of open sites). A feature like th
41.
by desterothx 1mo ago
Also, I can talk at a lot more than 3 tok/sec, its just going to be more gibberish (see thinking traces)
42.
by desterothx 1mo ago
The trouble is the article is so poorly written I don't want to look for hidden information, i want to close out. I'm guessing ai wrote it, the information is constantly repeating, and not even always consistently
43.
by desterothx 1mo ago
I actually noticed this too with DS flash over openrouter. Instructions to respond in my native language cause the model to spew gibberish roughly resembling my native language mixed up with similar languages, and it was also dropping in Cy
44.
by desterothx 1mo ago
I mean, it implies that it has an even better cache hit rate that DS flash, which is impressive, as the chr on DS flash was already really good in my experience
45.
by desterothx 1mo ago
the fee goes per product category, not per item...
46.
by desterothx 2mo ago
Last section?! You think people actually read these before commenting?!
47.
by desterothx 2mo ago
yes, hope this answers the loaded question :)
48.
by desterothx 2mo ago
yeah thats kind of the conclusion if you read the article entirely...
49.
by desterothx 2mo ago
if you have a legitimate issue just flag the post?
50.
by desterothx 2mo ago
bold of you to assume the ai companies care about the ethicality of their suggestions
51.
by desterothx 2mo ago
i mean i agree diversity of opinion is good for problem solving, but the comment you were replying to is saying telling people to think outside the box doesnt really do anything. The e.g. also seems really cherry picked, im sure after a 100
52.
by desterothx 2mo ago
yes, except experts are capable of saying I don't know, while LLMs will rather give any answer than do so most of the time
53.
by desterothx 2mo ago
i would assume internally they have even better tps
54.
by desterothx 2mo ago
at that price, probably loaded into ddr5 ram, a 3 bit or less quantization, and 3 tps
55.
by desterothx 2mo ago
The large models whose thinking traces are useful are safeguarded against this reasoning replaying. the small models are just designed for speed and efficiency, so these safeguards are a lot meaker, making the attack possible
56.
by desterothx 2mo ago
I love how some of the biggest advancements in llms came from the Chinese labs, yet people still jump to distillation being unreasonably effective. Distillation is very good at creating smaller models from large ones sure, but nothing to me
57.
by desterothx 2mo ago
most devs i talk to say that they would love to go back to coding without ai, if everyone had to do it, so there's that. Oh also scammers have gotten so much better with llms, so a significant amount of peoples lives are worse due to t
58.
by desterothx 2mo ago
sure they do if it makes them money, probably just not worth the controversy right now
59.
by desterothx 2mo ago
so you decided to vibe code a sandbox because windows sandboxing is lacking, instead of moving to linux where sandboxing is robust?
60.
by desterothx 2mo ago
not really, compiling isnt a heuristic problem, it has a lot less randomness involved

More ›