beltsazar

10 ms·

1.
by beltsazar 1mo ago
As a comparison, Qwen3.6 27B scores 38, which was the highest in its small model category (4B–40B). Qwen3.8 27B beats all medium models (40B–150B). It has the same score as DeepSeek V4 Flash 0731, which ranks #5 in large model category (&gt
2.
by beltsazar 1mo ago
Using a better quant Q6_K (Unsloth's) compared to your quant Q4_K_M (LM Studio's) yields this: https://tools.simonwillison.net/markdown-svg-renderer#url=ht... Chains exist. Red scarf is proactively added. ("M
3.
by beltsazar 1mo ago
Is this a new rule? I've seen some popular blog posts reposted here several times and still highly upvoted.
4.
by beltsazar 1mo ago
> To me the severity has nothing to do with how popular is a code path, but whether that code path is accessible to an attacker. That's true for targeted attacks. For untargeted attacks, attackers have little incentive to do the exp
5.
by beltsazar 1mo ago
> I’ve run 3.6-27B and 3.6-35B on 32GB locally for a lot of bulk non-code tasks. Do you mind sharing your use cases?
6.
by beltsazar 2y ago
The anthropic principle is ridiculous. Suppose that, against all odds, you survive the worst plane crash in history. Then you ask NTSB what caused the crash and why you survived. They answer: "Nonsense! You wouldn’t have asked the ques
7.
by beltsazar 2y ago
Fine tuning for the earth might be able to be explained away most easily, like you said. Fine tuning for the universe, though... Firstly, we have zero evidence for multiverse. Some scientists even argue that the idea is untestable and unfal
8.
by beltsazar 2y ago
This is a such common logical fallacy that we should have name for it. > No matter how low the odds are, the counts of those potential interactions bring this outcome to a certainty. "No matter how low", really? Are you suggest
9.
by beltsazar 2y ago
You're asking as if the other candidate is a no-brainer choice. If the other candidate were Kennedy, then sure—but they were not. In this case, many would be undecided and would vote not the best candidate, but the least bad one.
10.
by beltsazar 2y ago
Oct 7 happened and you're suggesting a different approach than a war, i.e. diplomatic solutions? That's too naive—not even the most pacifist country would do that. And let's not pretend that no diplomatic solutions have been
11.
by beltsazar 2y ago
But if they hide amongst civilians and Israel is too weak, what do you suggest Israel should do instead?
12.
by beltsazar 2y ago
Yes, cosmological principle is probably the most fundamental assumption in astronomy. Most people don't realize that science—and even everything in life—has to start from some axioms/assumptions, just like math. I first realized t
13.
by beltsazar 2y ago
> There's a good argument for immutability by default, but many programmers dislike all the extra declarations required. That's one little reason why Rust is loved by many: immutability by default. Meanwhile, it's not even
14.
by beltsazar 2y ago
> there are so many pitfalls when designing this Agree. When people hear the adage "don't roll your own crypto", they often think it refers to crypto primitives only. In reality, it's also hard to design a secure cryp
15.
by beltsazar 2y ago
> The costs associated with implementing this requirement can be very small, if not trivial. This is too naive. While it may be the case for single player games that use online connection only as a DRM mechanism (Hitman 3, Gran Turismo 7
16.
by beltsazar 2y ago
I think it aims mainly at games that use an online connection as a DRM mechanism, such as Hitman 3 and Gran Turismo 7. Both games are single player games that shouldn't necessitate online connection in the first place.
17.
by beltsazar 2y ago
But WAL is not for durability, but for atomicity and consistency. And yes, you need to use fsync to ensure durability.
18.
by beltsazar 2y ago
If ice didn't float, life as we know it wouldn't be possible anywhere in the universe. (It doesn't exclude life that doesn't require water, though—if that's possible.)
19.
by beltsazar 2y ago
> You can’t exclude non English languages being you would still be surprised if it was in Spanish etc. I'm not saying that the only surprising result is an English esssay. But sure, let's add all languages in the world. Getting
20.
by beltsazar 2y ago
> Similarly finding any shape in a random set of points is much more likely than the odds of any one shape. Obviously. But that’s not the point (no pun intended). My point is that most of the "shapes" would be just an unstructu
21.
by beltsazar 2y ago
I agree to almost all your points from the previous four comments, and I think so do you to my comments (because you didn't argue against my statements). We differs only on what to discuss. Before I give up on this discussion that&#x27
22.
by beltsazar 2y ago
> Sure, but given a large enough sample both will likely exist. This applies to every event with nonzero probabilities. What's your point? > Humans have been finding patterns in clouds, stars and even toast since time immemorial.
23.
by beltsazar 2y ago
> Again I think you are confused. Assuming random distribution, 'We Are Coming' is just as likely as any other similarly long structure to form. You are confused. How could we be back to square one? We've discussed it be
24.
by beltsazar 2y ago
> However that doesn't mean that a sphere is any more or less likely than any specific other structure. A shape/structure doesn't have an intrinsic probability. Your sentence is underspecified. Shape of what under what pro
25.
by beltsazar 2y ago
But of all perspectives in 3D space, there are only a fraction of perspectives that see it as a line. Most other perspectives see it as a circle/ellipse. So, the earth's perspective is not that unique—in fact, it's the most c
26.
by beltsazar 2y ago
Of course each instance has the same probability. But we're not talking about the probability of an instance, but rather that of a set of instances. In the dice example, it's obvious that the probability of getting at least one di
27.
by beltsazar 2y ago
> Finding ~50 dots arranged in a (very loosely defined) circle, from any projection, of a dense set of 2 trillion of them is very plausible. It depends on the size of the circle, though. The smaller the size, the more likely the probabil
28.
by beltsazar 2y ago
> The chance of some set of galaxies among all the billions in the sky doing this is much higher. Of course in relative terms it's much higher, but it doesn't matter—what matters is the absolute value. 10^-100 is much larger th
29.
by beltsazar 2y ago
> Random processes can appear to have meaningful structure, but that’s just because we value some outcomes more than others. No. It's because some structures are much much much less likely to form randomly than other structures. I
30.
by beltsazar 2y ago
> To me it feels very similar to an interface (trait) implemented by a bunch of classes (structs) Then you might not fully grok sum types yet. > From my perspective a discriminant vs a vtable pointer is a boring implementation detail

More ›