anthonypasq

4 ms·

31.
by anthonypasq 1mo ago
i just gotta say that i find it hard to believe that this comment is true for you but you hang out on hackernews. almost everyone here is a software engineer or works in tech, they all use ai all day to do their entire jobs. > even thoug
32.
by anthonypasq 1mo ago
yeah its called web search
33.
by anthonypasq 1mo ago
at this point, s3 just means object storage and doesnt mean it actually has to be hosted on aws. theres plenty of other companies that provide s3 compatible storage apis.
34.
by anthonypasq 1mo ago
isnt the active parameter count more relevant than the total? qwen is a dense model no?
35.
by anthonypasq 1mo ago
sure, but my point is that supply and demand are what determines pricing, not cost to serve. its like thinking that because something costs $1 to make, its not possible for a company to sell it for $10, or that becuase they are selling some
36.
by anthonypasq 1mo ago
no, because closed sourced model pricing has no relationship to its size. Thats what im saying. the inference margins are crazy, but people think the fonrtiner models must be 10T params or something because theyre expensive
37.
by anthonypasq 1mo ago
I'd just like to point out that the largest model Cerebras has ever served is Kimi K2.6 which is 1T parameters, so that either means that theyve had a breakthrough on the hardware engineering side of things, or GPT-5.6 Sol is likely a
38.
by anthonypasq 1mo ago
no its not. there is nothing in the code that can tell you how the system was intended the work. The existence of a bug is by definition related to information that is not contained in the code. it is a mismatch between intent and what the
39.
by anthonypasq 1mo ago
throughout history, Google has been obsessed with speed as a feature. that was a huge reason people used google search, and then chrome in the first place, and it think its really underestimated by people. Jeff Dean specifcally seems to thi
40.
by anthonypasq 1mo ago
is presumes you are doing longer difficult agentic tasks, if youre doing a simple problem in 1 or 2 shots, not really multi turn then theres no comparison.
41.
by anthonypasq 1mo ago
for non-coding applications, i think speed is a real differentiator. Im building an app that uses LLMs for some functionality that the user would not have any reason to expect is using AI and therefore having then wait seconds or minutes is
42.
by anthonypasq 1mo ago
flash-lite is more of their luna tier competitor but even still not quite there yet, but gemini's dominance on multimodal and image understanding i think really gets downplayed on this site when most people think the only think you can
43.
by anthonypasq 2mo ago
surely you could imagine that being able to sit in on a meeting that was had when they were designing the system would be useful for you? theres a lot of context and decision making that is useful but not documented in the code. The code te
44.
by anthonypasq 2mo ago
compilation time is massively important for developing with agents
45.
by anthonypasq 2mo ago
very interesting idea. i didnt think of that. i was just assuming youd have an additional one of these in your phone for actual lightning fast local inference
46.
by anthonypasq 2mo ago
im assuming energy expenditure is substantially lower as well
47.
by anthonypasq 2mo ago
Personally I think Apple should have acquired them. if you could burn a gemma4 class model into an iphone and actually get extremely low latency and low battery usage it would feel like the future IMO. even if it means you wont get frontier
48.
by anthonypasq 2mo ago
> wondering exactly how hard would it have been if they were to hire artist(s) approximately infinitely more hard, considering using AI is 0 hard.
49.
by anthonypasq 2mo ago
> Let's suppose each models was subsidized at 70%, so that we only pay 30% of the cost. why on earth would you suppose that?
50.
by anthonypasq 2mo ago
how many times do you have to be metaphorically hit in the head with a brick before you realize inference margins at api pricing were 80%+
51.
by anthonypasq 2mo ago
i wonder if Apple will eventually ship proprietary models with burned into the silicon for all local workloads https://eu.36kr.com/en/p/3904844399445638
52.
by anthonypasq 2mo ago
Im sorry but this is nonsense. You dont give people unrestricted access to explosives and then punish them after they blow something up. You prevent them from getting it in the first place if you actually believe the item is as dangerous as
53.
by anthonypasq 2mo ago
They will never do this because it will reveal way more about their other model architectures than they would ever be willing to do. I mean really why the hell are we mad at a company for not wanting to open source their IP? do you think Me
54.
by anthonypasq 2mo ago
unless you think that Opus is 10T+ params, its pretty much impossible for inference not to be profitable when doing some basic napkin math on other open models, and if Kimi K3 is 3T params with the same performance as Opus then that means t
55.
by anthonypasq 2mo ago
gen z and even more so gen alpha is a complete bimodal distribution. The top kids are genuinely incredible.
56.
by anthonypasq 2mo ago
> That is a paper loss and you have years worth of cash you can spend while waiting for it to recover. not if you're 80 dude...
57.
by anthonypasq 2mo ago
i dont think you want to ever be in a position where you arent making money and your net worth could drop 50% in a year. but hey, if you want the risk go for it i guess.
58.
by anthonypasq 2mo ago
no one that needs to rely on their investments for their actual retirement still has them in equities. theres a reason target date funds automatically adjust asset allocation as it nears its target date. you should be in majority bonds and
59.
by anthonypasq 2mo ago
retirees arent suppose to have their active retirement funds in stocks dude. Any financial advisor with a brain would not make such a ridiculous asset allocation error.
60.
by anthonypasq 2mo ago
> And neither do Lecun or Sutskever or Sutton! They are all focused on human intelligence. None of them are even slightly concerned about an AI which is intelligent before it learns any language. ??? https://www.youtube.com&#x

More ›