colingauvin

5 ms·

1.
by colingauvin 3d ago
I used to laugh at those "this software may not be used in the development of nuclear weapons" lines.
2.
by colingauvin 6d ago
But they aren't really competitive for cost in that size class.
3.
by colingauvin 6d ago
It's not particularly good value if you are just comparing $ with no other context. Point is just that it's accessible and so now Anthropic and OpenAI need to make both a performance proposition and a value proposition.
4.
by colingauvin 6d ago
Not to discredit your opinion here, but will just point out I am also a PhD biochemist who has worked on viruses, so hopefully our credentials cancel. :)
5.
by colingauvin 6d ago
OpenAI just paused new subscriptions to their $200 plan. They are in a rock and a hard place. Obviously the Astras and Fables of the world are exponentially more expensive, but for...less than exponential returns. The question is whether th
6.
by colingauvin 6d ago
I am for more sick of reading people complain about it than I am seeing it. The page is a table. It was fine.
7.
by colingauvin 7d ago
I would never advocate they not be allowed to say it. But I also am sick of hearing wildly unrealistic takes.
8.
by colingauvin 7d ago
This number gets cited a lot but it's actually the excess mortality of post-European contact, and it's not content wide, it's specific to Mexico. It also includes climate change, famine, war, relocation, etc.
9.
by colingauvin 7d ago
Arguably yes. I would wager no one killed by an asteroid in the history of people killed by asteroids realized what was happening before they died.
10.
by colingauvin 7d ago
Most people do not understand how easy it would be to make anthrax. Technology is not the reason it basically is never done or deployed. AI cannot meaningfully move this needle. Let's look at it in an inverse way. Making a single drug
11.
by colingauvin 7d ago
Oh my God THANK YOU. I do not have the words how tired I am of hearing tech people and economists talk about biology.
12.
by colingauvin 7d ago
yawn Can we just think through the logistics of AI taking over the world? Like, we won't notice it using all our GPUs? Like it somehow would have enough context to take over the world? lol.
13.
by colingauvin 9d ago
Yes but you will lose 120 FPS/VRR. So for high end TVs not really ideal.
14.
by colingauvin 11d ago
I'm not trying to read your comment as whataboutism, but it did get me wondering: From a moral or ethical standpoint, if there is a gradient route from "very immoral to moral", but it involves first going from "very immo
15.
by colingauvin 12d ago
Early Fairphones were not necessarily designed for repairability. That is a principle that has emerged significantly in the last few iterations. The FP4 was the first that was really designed and marketed as fully modular from what I rememb
16.
by colingauvin 14d ago
Fascinating. I work in this space sort of, there's still a ton I don't understand. What is the legal space like? Are you at risk of being shut down on a dime if the FDA decides Lilly can meet demand? Are you able to compound small
17.
by colingauvin 15d ago
Since you posted this the Asus GX10 went from $4000 to $6000. https://eshop.asus.com/us/ascent-gx10.html
18.
by colingauvin 16d ago
GLM 5.3 Flash? Qwen 3.8 Flash Next? I believe those both are as good as the best Gemini, competitive with Terra.
19.
by colingauvin 19d ago
Quality of tokens matters almost as much. 30 is worth it on a meaningfully better model, though annoying. 50 is pretty much unnoticeable (good old 60 FPS). 80-100 is heaven (aka 120 FPS - once you get used to it, it does suck to go back). I
20.
by colingauvin 19d ago
I can't recommend LocalAI enough. I am thoroughly enjoying it in all aspects. Especially the memory management and ability to load and boot models to and from memory as needed, right from the clients that use them.
21.
by colingauvin 19d ago
Finally switched over to entirely local AI models (Qwen 3.8 Flash Next) and...for the first time in a year, I didn't know Claude was down. What a phenomenal feeling.
22.
by colingauvin 19d ago
In terms of pure tokens per dollar, absolutely not worth it. That said, when I bought my pair of Sparks, the best model I could run on it was GPT OSS 120B. That has an AA score of 24. Today, the best model I can run on them is GLM 5.3 Flash
23.
by colingauvin 19d ago
PWAs suck and when I have to use one I immediately go find something else.
24.
by colingauvin 20d ago
Nvidia literally releases quantization tools.
25.
by colingauvin 21d ago
A single Spark alone is not worth the price. You are paying $1000 just for networking equipment you aren't using. At 2x it starts to maybe become worth it if you don't want to deal with Apple. Outside of the newest Macs, I can&#x2
26.
by colingauvin 21d ago
Your point isn't lost on me, but a few other considerations: 1) Rates are theoretically discounted for GLM 5.3 Flash right now, by 50%. 2) Hardware costs have continued ascending with no sign of letting off, so it's unlikely that
27.
by colingauvin 21d ago
Who the hell cares when I can run it myself?
28.
by colingauvin 21d ago
That's 160GB-ish for Q4...how is 256 insufficient?
29.
by colingauvin 23d ago
I don't know anything about blood pressure, really. But I do know that where this guy's writing touches my own expertise (infectious diseases, pharmacology) - he is often very very wrong.
30.
by colingauvin 23d ago
I'm supposed to believe that DeepSeek distilled a 300B-1T param model with 150,000 requests? Lol.

More ›