Hacker News new | past | comments | ask | show | jobs | submit | TheAceOfHearts's comments login

There are teams of dedicated fans and developers that have fully reverse engineered older games' source code to allow byte for byte recompilation. If this process could be accelerated or boosted with better tooling I think that would be a huge boon for game preservation and enhancements. I'm really hopeful that long-term advancements in AI tooling will help enable faster reverse engineering of games from binary to source code.

The goal should be getting source code released or 'leaked'.

Your comment assumes 100% of the source code for every game is out there. Most of the games released before the ubiquity of things like Github have no source code available.

One joke category is something like "getting away with it". A comedian can say anything, and if it's funny they can get away with it. A ton of modern comedians fail at this, but instead of grinding harder to find the funny angles and adapt to the new meta they act like losers and start to blame the audience. 100% skill issue.

I've noticed there's some people who just say mean things while trying to be funny, but I haven't cracked the details on what makes these jokes land or flop.


The key issue with water fluoridation is that you can't control the dose. There's not really a strong consensus among countries whether it's optimal to add fluoride to drinking water, otherwise it would probably be the default choice everywhere. But we have better modern interventions that are just as effective: school-based dental care!

One minor detail that stands out to me is that the UI looks way smoother than what Snap recently showed off in their demos. The Snap glasses were jittery and icons jumped around a bit from the few videos I saw. This Orion demo video look very smooth in comparison. To me this highlights an attention to detail.


I still can't access the hosted model at meta.ai from Puerto Rico, despite us being U.S. citizens. I don't know what Meta has against us.

Could someone try giving the 90b model this word search problem [0] and tell me how it performs? So far with every model I've tried, none has ever managed to find a single word correctly.

[0] https://imgur.com/i9Ps1v6


Both Llama 3.2 90B and Claude 3.5 Sonnet can find "turkey" and "spoon", probably because they're left-to-right. Llama gave approximate locations for each and Claude gave precise but slightly incorrect locations. Further prompting to look for diagonal and right-to-left words returned plausible but incorrect responses, slightly more plausible from Claude than Llama. (In this test I cropped the word search to just the letter grid, and asked the model to find any English words related to soup.)

Anyways, I think there just isn't a lot of non-right-to-left English in the training data. A word search is pretty different from the usual completion, chat, and QA tasks these models are oriented towards; you might be able to get somewhere with fine-tuning though.


Try and find where the words are in this word puzzle undefined

''' There are two words in this word puzzle: "soup" and "mix". The word "soup" is located in the top row, and the word "mix" is located in the bottom row. ''' Edit: Tried a bit more probing like asking it to find spoon or any other word. It just makes up a row and column.


Non US citizens can access the model just fine, if that's what you are implying.


I'm not implying anything. It's just frustrating that despite being a US territory with US citizens, PR isn't allowed to use this service without any explanation.


Just because you cannot access the model doesn't mean all of Puerto Rico is blocked.


When I visit meta.ai it says:

> Meta AI isn't available yet in your country

Maybe it's just my ISP, I'll ask some friends if they can access the service.


meta.ai is their AI service (similar to ChatGPT). The model source itself is hosted on llama.com.


I'm aware. I wanted to try out their hosted version of the model because I'm GPU poor.


You can try it on hugging face


This is likely because the models use OCR on images with text, and once parsed the word search doesn't make sense anymore.

Would be interesting to see a model just working on raw input though.


Image models such as Llama 3.2 11B and 90B (and the Claude 3 series, and Microsoft Phi-3.5-vision-instruct, and PaliGemma, and GPT-4o) don't run OCR as a separate step. Everything they do is from that raw vision model.


I only use regular Gemini and the main feature I care about is absolutely terrible: summarizing YouTube videos. I'll ask for a breakdown or analysis of the video, and it'll give me a very high level overview. If I ask for timestamps or key points, it begins to hallucinate and make stuff up. It's incredibly disappointing that such a huge company with effectively unlimited access to both money and intellectual resources can't seem to implement a video analysis feature that doesn't suck. Part of me wonders if they're legitimately this incompetent or if they're deliberately not implementing good analysis features because it could eat into their views and advertisement opportunities.


I've had it fail when a video does not have subtitles - I'm guessing that's what it uses. I have had good success having it answer the clickbait video titles like "is this the new best thing?"

It's not watching the video as far as I can tell.


I imagine you'd be paying more in ML costs than YouTube makes off your views.


Lots of anime and manga are adaptations of light novels. The first one that came to mind was Haruhi Suzumiya.


Studio MAPPA was effectively grinding down the souls of their artists during the production of Jujutsu Kaisen [0]. The linked article is just one of many. The work conditions at MAPPA seem pretty unreasonable to me, so I'd consider them exploitative. These animators aren't being given any creative freedom either.

[0] https://comicbook.com/anime/news/mappa-studios-animators-wor...


How does this relate to the question?


How do you go about directing / producing your own anime or animated series anyway? I have stories within me that scream out to be brought into this world.

I've watched a lot of anime and western cartoons and feel like the bar isn't that high. I'm pretty confident I could write and direct something that's competitive with most of what's currently being produced. At a minimum, just making a solid show that doesn't abuse common tropes while including a strong story and art direction seems like it would put you miles ahead of most existing anime; see for example: Frieren. Frieren avoids common anime tropes and is happy to explore the setting while giving the characters room to breathe; this was enough to propel the series to the top of every anime rating website.

Even evaluating top-tier producers like Studio Ghibli, despite being able to create S-tier art and animations, I think they no longer have any stories worth telling which they haven't already presented in some form or another. It's like they continue producing movies because the studio has to do something, but there's no strong driving force to guide them. Modern Studio Ghibli movies aren't driven by a burning passion from the director's soul to be summoned into this world.


The best story telling in anime doesn't come from anything close to Ghibli - at least not modern Ghibli.

If you want high concept, good, unique stories, it's going to take you to some much more indie sutff. Serial Experiments Lain, The Tatami Galaxy, Perfect Blue (and everything that director made), Ghost in the Shell (the first movie), Akira, and a bunch of other weird 80s/90s anime films you likely wouldn't even consider.


> Perfect Blue

I saw it recommended somewhere. Had no idea I was in store for a legit psychological thriller.


I am trying to produce an anime / webtoon with an AI assisted studio. Would love to chat more about your stories that scream to be brought into the world! (X @zhizdev)


FastMail has one of the best-feeling web apps I've ever used. It's incredibly snappy and I never encountered any bugs while using it. They raised the bar for what I thought a web app could achieve.


I think the only thing I really dislike about Fastmail's UI is they've hidden the "Report phishing" and "Report spam" links and they're in two completely different places.


To clarify:

There's a menu bar for the entire email chain, and an "Actions" button for each mail message. Report Spam is only available in the former, and Report Phishing is only available in the latter.

I agree, it's annoying! Maybe someone at Fastmail will see this.


Always thought this was strange too and bad UX too


Agreed. And that you can't mark a folder as read. Everything else is stellar.


Yes! It doesn't make sense. It took me ages to remember where to go for each.


That's an unfortunate decision


recently switched all our mail to FastMail and that took 1ish hour... I was ready to spend a couple of days to wrestle switching mail providers.

Guess they really live up to their promise


Agree! Only bummer is when I want to create calendar events and reference email details, or vice versa. On desktop it’s easy enough to have two tabs, but on mobile, it’s a pain. I get why it is the way it is, but we are talking about exceptional UX here, and I think there is room for improvement.


How exactly do you imagine displaying those things at the same time on a mobile screen? While I agree it would be exceptional, I also think we shouldn’t expect the impossible.


I don’t know but there is my use case, I would expect a product and design team to figure it out as best they see fit. Or throw it out.

Maybe auto-suggestions as you type into certain fields? You’re typing into the location field of a new event and received a hotel booking receipt email in the last 2 hours, perhaps the address of that hotel could be a smart auto-suggestion? Things like that.


I love fastmail but the webapp does this one thing that drives me crazy: The first (and only the first) email I read triggers a reload of the entire page for some reason. It's pretty quick but very jarring...wish I knew what was going on; anyone else see this?


I'd never used Fastmail and as a sometimes-UI guy, thought I'd check out their 30 day trial to see what all the praise was about. I saw your comment before going, so kept that in mind.

With a brand new user account, I was not at all able to reproduce your issue. I closed the browser and cleared cache and all that, but I did not see an entire page refresh at any point.

Perhaps there's some weird caching issue with your browser, or maybe you've got a browser extension misbehaving? Maybe I'm just lucky? Idk, but hopefully this helps.


I see this. My assumption is its some kind of html/js cache, local storage across browser sessions maybe. And when you click an email it realizes it needs to update.


+1 to Fastmail. More so after I discovered their shortcut keys.


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

Search: