vova_hn2

> Is it because the criteria is wait time not travel time?

Yeah, it doesn't make any sense to optimize wait time instead of total time from pressing the button to reaching destination.

If I'm writing a text about Russian language in English, I can easily include an example in Russian: "привет, как дела?" and it will be displayed correctly.

But, if I was writing about Mandarin Chinese in Japanese, I just wouldn't be able to do it using a simple text input (like HN comment form). I would need more advanced formatting tools to make sure that both Chinese and Japanese characters are displayed correctly.

Because, as far a I understand, one Unicode symbol could correspond to both a Hanzi character and a Kanji character. Which character will be displayed is only determined by font. Since HN doesn't allow me to select different fonts for different parts of my comment, it is simply impossible to create a comment that includes both Japanese and Chinese characters.

> But also, maybe consider not giving a fuck about what someone on the internet think?

If I'm writing something, typically I want to share my thoughts with other people and get their feedback. It won't work if they dismiss my writing because of some superficial stylistic choice, without even trying to engage with the ideas that I'm trying to express.

> What the hell are these people doing?

Goalpost shifting, mostly

I tried the game on mobile first and couldn't figure out the you are supposed to press "go". On my phone it looked just like all other buttons and was not highlighted in any way. Not sure if the reason is mobile browser or dark theme.

> it's just like 5 people doing most of it

I understand that finding technical/mathematical solutions to social problems is usually pointless, but still...

What if we have, like, a lot of reviewers.

Each reviewer gets n books randomly assigned to them.

n is picked in a way that allows each reviewer to thoroughly read each of them without dropping it in the middle for bullshit reasons.

Each reviewer ranks n books assigned to them, then we use an Elo-like algorithm to produce the final ratings and pick a winner.

This would eliminate most of the randomness/unfairness described in the article.

Of course, reviewers wouldn't like it, because it would significantly reduce their power and people never give it away willingly.

This is so called "Atomic" (rpm-ostree based) Fedora, so flatpak is the native way to install applications

> The Russian language itself comes from Old Bulgarian / Old Church Slavonic, as does Serbian and other "Slavic" languages.

Also, Bulgaria used to own all the land in the world, but because Bulgarians are very kind people, they gave some of it to other nations so that they have a place to live too. Thank you, Bulgarians!

I have 152.0 installed from flatpack on Fedora Kinoite 43 (Plasma 6.6.3 / Wayland)

Maybe it's a firefox bug that was introduced somewhere after 140? idk

Or you can just ignore the Earth's rotation completely and sleep whenever you want:

https://xkcd.com/320/ https://www.explainxkcd.com/wiki/index.php/320:_28-Hour_Day

I'm sorry to say, but in the both examples on the page I liked the original version more.

In the first example guitars sound very lively, like I'm actually listening to someone playing live. Very warm and cosy feeling. 2026 version is just some standard over-produced pop-music, that I would have completely ignored if I heard it "in the wild".

In the second example, 2003 vocal performance might be not perfect from a technique perspective, but I can hear real emotion in his voice. 2026 version sounds very bland, just like every other song that I might here on a pop-radio.

I want to say that I am not an anti-AI fanatic, I am generally curious about new ways to use technology (including gen-AI, but not limited to it) to create art. So I am not coming from "it's AI, therefore it's bad" perspective. I genuinely tried to listen with open mind and hear the music without thinking too much about how exactly it was created.

I can confirm, my firefox on linux crashes immediately as well.

Curiously, the website works just fine in chrome on android.

Blink monopoly strikes again, I guess.

Why not use UTC?

I find unequal hours [0] interesting.

Basically, day (from sunrise to sunset) and night (sunset to sunrise) are each divided into 12 equal periods. But night hours and day hours are, of course, not equal to each other and changing throughout the year.

You can make a computer implementation by pulling astronomical sunset/sunrise times for a specific geographical location and then it's a simple arithmetic to convert from modern hours to horae temporales

[0] https://en.wikipedia.org/wiki/Unequal_hours

Aren't modern months made up as well? They are not in sync with lunar cycles anymore.

> thanks to them we have satellites in the first place?

What kind of astronomy knowledge is required to launch a satellite?

> No UN body can command a nuclear sovereign.

Why not, though? If any country violates their limit, just issue a concern. If they ignore it, upgrade it to a grave concern. Then they will surely have to obey, I mean, it's grave concern we're talking about, what else they could do?

I don't like the title, it implies that the only reason for using an ORM is not knowing SQL, which is obviously not the case.

Every time I tried to do a project without an ORM, using only raw SQL, I inevitably ran into:

- serialization/deserialization boilerplate. Like, having to manually map values returned by the DB library to object (or named tuple, or structure) properties

- poor code reuse, having multiple very similar queries that have just one small difference

- extra pain in changing DB schema. Adding a field requires to go and manually edit many queries

Anti-ORM crowd never gives a good answer to these issues.

Instead, they push strawman attacks like "oh, you only use ORM, because you can't write raw SQL". I can absolutely assure you that this is not the case. Every time I use an ORM (SQLAlchemy mostly, the one mentioned in the article) I am 100% sure what SQL do I want it to produce and what SQL will a particular ORM invocation produce.

Sure, there is some overlap. But there are also certain traits that would mean instant elimination from the gene pool in the wild but don't matter much when a human takes care of the horse.

I'm sure that if you investigate feral horse populations like North American mustangs or Australian brambies, you will find out that they don't "get ill if they breathe wrong". Because all that did, died without leaving any offspring long time ago.

Selective breeding for being useful to humans, not for being able to survive on their own.

> LLMs have seen humans act like conscious beings all over their training data because humans acting like conscious beings IS their training data.

How do we know that humans don't learn how to act conscious by observing other humans who act conscious?

Consciousness doesn't have a precise definition, but if you ask someone to describe it, there is a good chance that the description will include the concept of internal monologue.

The problem is that "internal" monologue is completely meaningless if you never heard an external monologue.

Also, people usually describe internal monologue as something that uses language and language is impossible to learn without communicating with other humans or at least observing other humans.

What I'm saying is that "well, LLM just pretends to be conscious, because it observed humans acting like conscious beings" doesn't really helps us to create a meaningful distinction between human consciousness and machine "consciousness", because same can be argued about us.

We don't know if feral children [0] are conscious and we don't know how to check it.

[0] https://en.wikipedia.org/wiki/Feral_child

What would happen if the file was edited concurrently? Would any data be lost?

I wish it would show the whole chain in the end, perhaps augmented with dictionary meanings or links to Wiktionary.

The idea that they really send spam to validate an email address sounds to insane to be believable.

Is it possible that they are somehow leaking the address to actual spammers?

For example, they (or the hypothetical email validation SaaS) use an infected email validation library that ex-fills every email supplied to it, or something like this.

Role tags are not actual symbols "<system>", they are special tokens that do not correspond to any normal text. So you can't really inject a role tag, that is not the actual problem.

> I can distinguish my own thoughts from your speech without effort; they arrive through completely different channels with completely different sensory signatures. But for an LLM, everything arrives through the same channel as one long token soup. Its own thoughts sit next to your instructions, which sit next to the contents of a random webpage it just fetched.

I was thinking about the original encoder-decoder transformers, that did have separate channels for input and their own output.

Why can't we bring it back? For example, one channel for system prompt and another for everything else.

I cannot imagine git being a performance bottleneck in agentic workflow.

> You can work on many tasks in parallel without needing to download everything or fight worktrees.

What does "download everything" even mean? Why would you "fight worktrees"?

By the way, if you are interested in nonograms specifically, there is a great website nonograms.org that has tens of thousands nonograms (both B&W and color) and no ads.

Interestingly, the pixelization/noise effect is applied clientside, so if you open an image in a new tab, you can see the original. Originals look much better, in my opinion.

> A tangent: writing distractors for multiple choice questions is hard.

In case of online quiz you can have a "competition" between distractors:

1. start by having much more distractors than needed and pick randomly

2. for each measure the probability of it getting clicked (clicks/times it's shown)

3. show the most frequently clicked distractors more often