Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
cubefox
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
16 ms
·
451.
▲
by
cubefox
5mo ago
The other reason is lack of continual learning, especially for long games like RPGs.
452.
▲
by
cubefox
5mo ago
Song starts at 1:41.
453.
▲
by
cubefox
5mo ago
There is also strong evidence that the shingles (herpes zoster) vaccine reduces dementia risk: https://arstechnica.com/health/2026/02/could-a-vaccine-preve... Herpes simplex also increases Alzheimer's ri
454.
▲
by
cubefox
5mo ago
They don't necessarily need to have polygon faces when you zoom in, they can just be rendered as dots with a single color, as long as adjacent voxels don't show gaps (unlike point clouds). Similar to how there are different metho
455.
▲
by
cubefox
5mo ago
In the present case ( Comanche ) they clearly do represent integer heights.
456.
▲
by
cubefox
5mo ago
> It's kind of weird to call them "columns of voxels" when the columns can't have gaps No, it's not weird. The columns don't have gaps because they are columns represented by a height map, which can't d
457.
▲
by
cubefox
5mo ago
No? Each pixel on a height map corresponds to a column of voxels of the specified height. You could represent the same height data with a fully general octree and it would look exactly the same.
458.
▲
by
cubefox
5mo ago
I liked it. You can tap on them and they hop.
459.
▲
by
cubefox
5mo ago
Let's say it like this: at least being on the "autism spectrum" impairs the reliability of various gut intuitions.
460.
▲
by
cubefox
5mo ago
Note to readers: I had very bad results with trusting my gut.
461.
▲
by
cubefox
5mo ago
What happened to adversarial attacks? I.e. noise that makes an image look like something else to a classifier than to humans. I guess frontier LLMs are no longer vulnerable to those?
462.
▲
by
cubefox
5mo ago
> Certainly not on the US We don't know how much higher the drug usage would be if all drugs were legalized. > or any country where the criminalization is in full force. That's an extremely general statement which is almost
463.
▲
by
cubefox
5mo ago
The ""war"" on drugs seems to be pretty successful in some countries where illegal drug usage is quite low compared to others.
464.
▲
by
cubefox
5mo ago
Considering that there are a) many drugs that are substantially more dangerous than marijuana and that b) addiction does create users which are not in full control of their actions anymore, the term "drug victim" does make sense.
465.
▲
by
cubefox
5mo ago
Weed isn't harmless. For example, it can probably lead to long-term brain damage. It is also toxic for the lungs, although less so than tobacco, which is usually consumed in greater quantities.
466.
▲
by
cubefox
5mo ago
False. There is plenty of empirical data on the destructive effects of many drugs.
467.
▲
by
cubefox
5mo ago
Drugs are absolutely not victimless, except perhaps for caffeine.
468.
▲
by
cubefox
5mo ago
Exactly. Europe is not a country. Some countries in Europe are more different from each other than others are from the US.
469.
▲
by
cubefox
5mo ago
Several new DRAM fabrication plants are expected to come online in 2028: https://manufacturing.economictimes.indiatimes.com/news/hi-t...
470.
▲
by
cubefox
5mo ago
This already happens on Hacker News. People frequently flag comments they strongly disagree with, rather than just downvoting them. Despite the fact that the "flag" was originally introduced to flag spam.
471.
▲
by
cubefox
5mo ago
It's likely still a lot better than the current situation of not having any detection.
472.
▲
by
cubefox
5mo ago
This is great. We don't want to drown in AI slop, and (perhaps more importantly) we don't want people to think that real videos are AI generated. Any signal which helps distinguishing the two is helpful, even if it isn't perf
473.
▲
by
cubefox
5mo ago
Yep. Zero temperature is neither necessary nor sufficient for deterministic inference.
474.
▲
by
cubefox
5mo ago
The article didn't say that a pacifist run was more difficult than a genocide route.
475.
▲
by
cubefox
5mo ago
To clarify the lighting point in the article: there are two basic parts to realistic lighting, direct illumination and global illumination. Direct illumination displays only the "first hit" light that comes directly from light emm
476.
▲
by
cubefox
5mo ago
For a more thoughtful and very different analysis: "Implications Of Predicting The Next Token" https://minihf.com/posts/2026-05-07-implications-of-predicti...
477.
▲
by
cubefox
5mo ago
> I attempted to provide a number of different angles and evidence for the conclusion that the papal encyclical on safeguarding the human person in the age of AI is actually an article in large part neither on nor in AI, but by A
478.
▲
by
cubefox
5mo ago
> So my actionable question is "How do I convince my doctor to get the screening?" I guess we have to pay for it ourselves. It's not exactly cheap, but perhaps worth it.
479.
▲
Claude, Author of the Humanitas
(lesswrong.com)
2 points
by
cubefox
5mo ago
|
1 comments
480.
▲
by
cubefox
5mo ago
> Aren’t you already using English in the LLM convo? Telling the model to use French for research or to find resources in French seems like a reasonable step. Most ordinary people will just use their native language and they have no way
More ›