ActorNightly

5 ms·

61.
by ActorNightly 1mo ago
> Qwen3.8 27B (Q4_K_M, 17GB) generates at ~14 tokens/s on my Mac Studio M3 Ultra >~14 tokens/s For anyone reading that has never ran local llms, please understand that anything under 100 tok/sec is worthless. You are fa
62.
by ActorNightly 1mo ago
Trump won because of apathy. https://www.reddit.com/r/Infographics/comments/1gms8gb/the_2... The critisism isnt about the policy on immigration. The critisim is about analogously giving a convicted felo
63.
by ActorNightly 1mo ago
Yeah thats called xenophobia. It be valid point if you cound point to a set of traditions and beliefs that defines what American culture is, but modern American culture is literally an average over contributions from different cultures due
64.
by ActorNightly 1mo ago
Its also helpful to remind folk about why someone would hold this position. The largest reason US got to where it was at least in the world economy stage was because of the visa program. It was the country everyone wanted to move to, becaus
65.
by ActorNightly 1mo ago
hugging face is basically a real easy way to run llms. They also provide a bunch of libraries to do things like split compute across all your resources.
66.
by ActorNightly 1mo ago
Name one feature that Mac has that you think is so great.
67.
by ActorNightly 1mo ago
Except there isn't. Or they never make it to top. And no, being nice to people that have no problem just straight up lying about shit is not the solution.
68.
by ActorNightly 1mo ago
Running agentic loops doesnt mean just being able to execute them. When your llm is so slow that you are faster writing the code yourself with free gemini that comes with google account, local llm is no longer worth it. Tok/sec is ever
69.
by ActorNightly 1mo ago
... untill Musk no longer matters. After that, that project is going in the trash faster than you can blink. Its a collosal waste of time, money, and resources, and its never going to work.
70.
by ActorNightly 1mo ago
GFX vram is still faster.
71.
by ActorNightly 1mo ago
Running local models isn't about the cost, its about owning your own data and being able to run uncensored models. If you just want to code with an assistant, local inference is by far not worth it. You also don't need to run the
72.
by ActorNightly 1mo ago
I dunno why Apple fans do this thing where they pretend that their specific workflow is the standard for how things work, and because they can do it so well on their Macs, that means Macs are the best. To be specific, 5 models at 57 gb mean
73.
by ActorNightly 1mo ago
> Perhaps in a few years when both local-running LLMs are good enough on regular consumer hardware there'll be a good reason to upgrade. I mean, local LLMs already are really good if you own Nvidia Cards, not sure what you are waiti
74.
by ActorNightly 1mo ago
Its bad, but in the end nobody really cares that much about what is good or bad.
75.
by ActorNightly 1mo ago
Honest question - why are you so stuck on Macs for local inference? A 4 GPU linux box with 3090s, which are $1500 a piece right now, will blow this thing out of the water. Even 2x3090 rig will run most of the good local models like Gemma4:3
76.
by ActorNightly 2mo ago
Lol think about what you are saying. You are basically demonstrating the non-intelligent people. If someone ego gets in the way of understanding something, or they don't pay attention to counter evidence to what they believe in, that i
77.
by ActorNightly 2mo ago
> but behind the scenes it powers everything from blogs to logistics to networking equipment to SIM cards. All because of baked in legacy. This is going to change in the next few years. You beat Java with Python today very easily - write
78.
by ActorNightly 2mo ago
Its not wrong, its just that people don't see the writing on the wall. It was the same conversations about Haskell, how functional programming with no side effects was the future, e.t.c and so on. There is literally ZERO reason to use
79.
by ActorNightly 2mo ago
>It is strange that this chip is so crazy efficient though. Its only strange if you believe Apples lies. The truth is, architecture doesn't matter. Single core performance, Apple was always on par with the best chips. Compute is com
80.
by ActorNightly 2mo ago
Oracle Java is also basically effectively in KTLO mode. The biggest use of Java currently is Android, and Google is on a path to migrate to Fuchsia eventually. The second biggest use is Apache Spark/Kafka, but with LLMs now, you can pr
81.
by ActorNightly 2mo ago
There is basically zero chance of that happening. Not that this is impossible to do, but under Elons leadership it will never happen. Intelligence is non separable, you can't be smart in one area and be idiotic in another - if you are
82.
by ActorNightly 2mo ago
In my case, its reversed. Amazon does more good than harm.
83.
by ActorNightly 2mo ago
The best thing to do in such a case is issue the following statement - "I am in full physical compliance and will not interfere with investigation in any way, but I plead the 5th on answering any questions, and plead the 4th on any per
84.
by ActorNightly 2mo ago
The issue is that once you are a person of interest with a real id, you become a target. The way to outsmart them is to stay maximally anonymous.
85.
by ActorNightly 2mo ago
No the company that donates millions to education, support programs, and improving quality of life for everyone.
86.
by ActorNightly 2mo ago
Having met Jeff Bezos personally once (ex Amazon employee), he seemed very down to earth. His mannerisms were almost awkward nerdy guy like, you definitely didn't feel like he was above you in any form and way. I dunno if that changed
87.
by ActorNightly 2mo ago
to call him an innovator is a slap in the face of everyone who is an actual innovator. The only thing he was good at as lying to create a brand image.
88.
by ActorNightly 2mo ago
In places like China, its really not that unthinkable to basically raise kids indoctrinated into an ideology and train them in the necessary skills so that you have a cyber army at your command. Also >Frequently they may not care about a
89.
by ActorNightly 2mo ago
It lowers the economic cost of performing an attack in the same way a gun lowers the economic cost of killing a person. It does nothing for consequences of that attack.
90.
by ActorNightly 2mo ago
Its like you don't even read the post or the content. Hint: agentic loops. No its not a prompt away.

More ›