LLMs do solve the discoverability problem which underpins every power user feature ever. It is fantastic to be able to describe what you want without the domain knowledge that was previously a prerequisite to achieving that goal. Hopefully as more systems create declarative ways to interact with them (MCP, clis), power users will be able to profit from this as a side effect as well.
You think this topic should be banned because you don't agree with the people discussing it? I actually think there's a good argument here that the topic is politics, which is off-topic and doesn't belong on HN. But wanting it banned because you disagree is just childish... You also added nothing of your own to the discussion, simply stating that everyone else is ignorant and prejudiced.
I don't know much about this, but when I visited Japan, I walked aimlessly in the outskirts of Tokyo (Sumida city) until I ended up in a homeless area. They had used old palettes and tarps to build square houses with functioning doors, and it was surprisingly clean. Their shoes were neatly lined up against the walls, they used trash bags and the ground was clean. It was quite a culture shock, because it was cleaner than non-homeless streets at home.
Layout, colors, font and especially language all immediately scream AI. It's fascinating how many of these articles make it to the front page. I would think almost everyone here would be an expert at spotting AI generated slop by now, but people keep up-voting it.
Even cheap quartz movements tend to drift by ±10 seconds per year, so you'd only really set it at dst. Maybe you're thinking about automatic watches which can drift by several seconds per day, and stop if you don't use it for 24/48/X amount hours. I have to set my automatic quite frequently
I agree with you, but I just want to point out that there are other seamless solutions to memory management like refcounting, used by for example GDscript (Swift? Perl?). You definitely wouldn't consider this manual memory management
I'm not so annoyed by it except that I liked the old design (smaller text) better. What puzzles me more is why it takes >1000ms to open a simple right click menu? And even after it opening, it often has to "Load" some options? In what world does it take over a second to open a simple popup on modern hardware?
Remember when stable diffusion would create image soups full of horrifying humanoid faces and figures? I imagine something similar happens when you dull the brain in the right way.
Is the human brain capable of doing real simulations of light and physics in realtime? Or does it hallucinate the details and represent some low-resolution mental image? You may be overestimating the capabilities of flesh-based neural networks and underestimating the capabilities of silicon-based neural networks.
To be fair, a lot of programs rely on those standard directories from the "Known Folders" specification - it would be hard to rip them out without breaking a lot of software. Even on linux it's hard to get away from these directories, as virtually every distro ships with xdg-user-dirs. Even if you use Arch, you'll probably install xdg-user-dirs with Gnome or KDE which creates those directories on login.
All popular social media platforms today (snapchat, instagram, facebook) have endless streams of unfiltered short-form media which is designed to addict. I think few people are actually criticizing online communication for kids (though there are concerns about what communities and people they may interact with). The major problem is addicting media which promotes asocial behavior and sucks away the joy and creativity of going through life.
Following this logic in context of GP's hypothetical, your son would end up taking heroin with his classmates covertly because he knows his father doesn't approve of it...
The point that (I think) GP was making is that it's impossible to raise kids in a different world than their peers. If their peers all have phones, your kid gets left out if he doesn't have a phone. If society bans smart-phones for kids under 15, there's nothing to be left out of since none of them have it.
IQ4 qwen 122b-a10b would mean 61GB total size and 5GB active, so about 5GB of the model loaded into GPURAM plus any generated context, and 61GB of weights loaded into system RAM? I don't know if that math is correct, but does that run well? Wouldn't that only leave 3GB of system RAM?
Nathan is right though. It specifies a Habsburg jaw, not that the frog is a Habsburg king/prince. It's understandable that the LLM will hallucinate a king from this, but it isn't what's being asked of it.
My experience with CRUD backends is that you get bogged down in really convoluted business logic and abstraction layers, and most of your time is neither spent writing or thinking, but debugging. (Of course debugging involves thinking, but it requires you to step around, print, inspect, and explore how data flows, rather than exclusively thinking up new solutions). And in my opinion, debugging is easier if you can faster navigate your code and more quickly type or edit small changes.
Could you expand on why you think Christianity made Hebrew/Judean culture less monotheistic?
Also, surely you would concede that scholars who sought to rediscover ancient greek/roman literature through a Christian lens in the 15th/16th centuries played a big role in making it accessible to the Christian world?
I don't think the blogger uses compromise to refer to conflicting personal opinions. Consider a complex UI like photoshop. If you have too many buttons available on the screen at all times, the UI becomes bloated and hard to navigate (cough, GIMP). If you cut out too many features, you'll loose users who rely on those tools. If you move items to sub menus, or sub sub menus, you make the UI cleaner, but risk users never finding those features, or you risk introducing too much friction for anyone to want to use them. So the compromise here is what features to present the user for quick access, and what features to hide behind indirection. This is a difficult problem, and requires deep understanding of your users and their needs/habits.
I've seen a few people on HackerNews swear by Fossil ( https://en.wikipedia.org/wiki/Fossil_(software) ) which uses SQLite as the backend instead of the filesystem. There are a few other approaches as well
Let's hope Valorant, Call of Duty, Battlefield, EA Sports FC and others who explicitly deny their anti cheat from working on Linux react to the increasing demand for gaming on Linux, and begin supporting anti cheat on Linux.
Also, if his friends play Halo, CS2, Ark, War Thunder, Arma, Hell let Loose, Splitgate, TF2, DOTA, Overwatch, or The Finals, he'll be fine.
I'm curious which ones you had trouble with? I switched last week, and have been testing all the games in my library. I'm quite impressed with how all of them just work with zero or little tinkering. The only games I haven't been able to play are the ones with anti cheat that explicitly deny Linux.
That is confusing. The Nintendo switch is half the price, and plays all of those games. Why would someone want to buy the more expensive Steam machine just to play 2d pixel art games?
I didn't criticize either your article or your comment :). I enjoyed reading the article and finally learning why it's called godan and ichidan, and it was fun to learn that kau was originally kawu.
Any mention of Japanese learning always brings out negativity and criticism (and the effect is probably doubled by being on HackerNews) so I understand that you're on defense.
My comment was really just an objection to "You still have to learn that and understand that", as I don't think studying grammar is a mandatory step in learning a language (I'm a subscriber to the Input Hypothesis by Stephen Krashen). Though it could very well be that grammar study is an effective shortcut to internalizing grammar. At the end of the day, the amount of hours spent learning from context matters way more than the specific steps taken along the way.