Hacker Newsnew | past | comments | ask | show | jobs | submit | abhpro's commentslogin

The author is more pointing out that these games don't look realistic. Look at the foreground of the HZD shot - why is it almost black in daylight?

To be fair - if I remember the location correctly - that screenshot is somewhat misleading because it's camera position is from the inside of a large ruin, with the ceiling and right wall of the "cave entrance" being just outside the frame.

Zelda looks realistic to them?

No, the author posits that Zelda explicitly goes for artistry and ignores any pretense of realism (that then falls flat on it's face when using an over-contrasting tone-map like in the HZD screenshot).

The problem I personally have with the Zelda example given is that it looks really bland to me - the landscape looks really washed out - the author says "Somebody would paint this. It’s artistic.", but I don't think anyone would paint with such bleached-out colours.


In the painting there's a delicate interrelation between colours - you have browns/greens/blues in the dark parts, and more whites/yellows/blues/pinks in the light parts. I wouldn't describe it as bland, though it is in a sense washed-out. BotW doesn't, and probably can't have that level of handling of shades colour in the enviroment graphics if nothign else because of the technical constraints of the Switch hardware.

Compare, if you will the sunset shown here from BotW: https://imgur.com/a/TvS2iHa with the 'hand'-drawn art from the switch store page: https://imgur.com/5clZZjg

Looking at the screenshot, what can you say - you can say that it's nice that the green/yellow of the sky is mirrored in the green landcape with yellow rivers. And the back-lighting of the sun is helping give definition to some of the mountains/hills, which is nice. But I don't see very much subtle going on with the landscape.

Looking at the art, you can see a lot more dynamic range, clearer silhouetting of mountain ranges at various distances, whereas the actual game is more monotone-green. You can also see the fog doing a lot more work of making the shape of the land clear. There are some bits of fog/mist in the screenshot as well, but they're not doing as much heavy lifting in terms of giving shape to the landscape.

The Switch is really limited on the hardware front, and I can't imagine what kind of trade-offs the art team had to make to get to where they are - it's a very difficult balancing act that I only understand a small part of. Nintendo also tend to be very conservative/restrained in their 3d style (I remember being somewhat unnerved by Ubisoft's "Mario + Rabbids Kingdom Battle" Mario game, because it went super high-production-quality).

It feels a bit cheap to give as an example, but the 3D MMO Love by eskil steenberg tried to emulate the impressionist style, and did a striking job: https://imgur.com/9U18eRZ The bloom effect is doing a lot of heavy lifting to make the bright colours pop, but even in the less glowy areas there is quite subtle layering of colours going on and one does have the feeling that the colours are playing with eachother. ( https://www.youtube.com/watch?v=Cc02ijaw-Tg a video of it in action, if you are curious ).

As another comparison, looking at elden ring you can see they've gotten 'using fog to make landscape silhouettes pop' down to a fine-art (maybe they're even over-reliant on it) https://imgur.com/a/5GEePwL And looking at the landscape you have really nice looking brown/oranges in the fields in the foreground, black/greys/browns in the mid-ground, rocky cliffs, fog is actually glowing, and you have some green forests in the top-left. That's a lot of nuance for what's essentially a brown landscape. BoTW doesn't have that - would it have it if the team had the hardware capabilities and time and budget? Who knows...


are you familiar with any famous paintings almost at all?

Oh, I see. I disagree that the original HZD had a pretense of realism though. The remastered version does and well illustrates the uncanny-ness https://www.youtube.com/watch?v=IlWK_ELBW08 . The outrageous god rays, bloom and lens flare in the remaster compensate for that because you can't actually see anything due to them blinding you...

I assume stuff like this is why some languages or script can never be deciphered


Well there will always be an audience that wants to hear what humans can mechanically pull off


Also chiming in to say you're wrong, I mean they're correct


I just think it looks ugly, and I don't want to create a product that I don't personally like. Totally open to accessibility and using existing frameworks but not Material.


Damn you're worked up. Some of my gripes with Windows come down to peripherals, where I'll spend a lot of time troubleshooting why my bluetooth device or speakers or mic don't work. There also seems to be no way to bypass using a password or PIN on startup without changing the registry. I'd like my computer to just stay on at all times so I can remotely connect to it, but what do you know, a forced update caused it to restart, and because it requires a passeord to get to the desktop I have no way of getting Parsec to connect. Yeah I tried to disable automatic updates but nothing seems to stick. Why is the mic on my PS5 controller connecting and disconnecting ten times a second. Ok let me just try to unpair the controller, oh it just... won't unpair.


Are the ps5 controllers supported without having to buy software? If not, then I sarcastically say "I'm shocked that a competitor's peripheral works poorly on their product"

Because why wouldn't it work poorly?


Source on hallucination rates falling?

I use LLMs 20-30 times a day and while it feels invaluable for personal use where I can interpret the responses at my own discretion, they still hallucinate enough and have enough lapses in logic where I would never feel confident incorporating them into some critical system.


My own experience, but if you insist

https://www.visualcapitalist.com/ranked-ai-models-with-the-l...

99% of systems aren't critical and human validation is sufficient. My own use case, it is enough to replace plenty of hours of human labour.


Dark souls for sure now


I don't understand why automated applications is unethical. If most companies won't spend the time to manually look at each application, why expect applicants to manually apply to every company. I don't even use services like those but I think the current system is flawed for today's job market so it's hard to even say what is ethical right now.


I already said: we review every application manually, as do most people who have spoken up on threads like this. All you're accomplishing is burning out those of us in hiring, which makes the market worse for everyone.

It's unethical by Kant's universalizability principle: if everyone did it the hiring market would turn into a complete craps shoot.


> All you're accomplishing is burning out those of us in hiring, which makes the market worse for everyone.

The fundamental difference of course being that the applicant might need the job to eat and put a roof over their head, while the hiring people are employed and doing their job.

I get it, it sucks to be burnt out. But the often forgotten nuance of the job market is that there is a huge power asymmetry in hiring.


And I'm emphasizing that this kind of behavior is not going to get you a job, it's only making everyone's life worse. It's counterproductive in the extreme.

Give me one testimonial of someone who credibly demonstrates that these applications work.


>I already said: we review every application manually,

I would say your company is a rarity.


On what basis? The only people I've seen claiming this is prevalent are frustrated applicants, but I'd love to hear from someone who actually has seen the inside of one of these companies that uses AI filters in bulk.


It's a chicken and egg problem. Companies would likely look at every application if there were only a few. But if there are 1000 for every position and most are automated then companies have to automate filtering on their end too.


But is it better than Claude?


I generally find o1, or the previous o1-preview to perform better than Claude 3.5 Sonnet in complex reasonings, new Sonnet is more on-par with o1-mini in my experience.

Would expect o1-pro to perform even better.


Genuinely curious to know. Nothing I’ve used comes close to Claude so far.


Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: