Hacker Newsnew | past | comments | ask | show | jobs | submit | Lapel2742's commentslogin

> I'd skip this for now - it does not allow any kind of interactive conversation - as I learned after downloading 5G of models - it's a proof of concept that takes a wav file in.

I haven't looked into it that much but to my understanding a) You just need an audio buffer and b) Thye seem to support streaming (or at least it's planed)

> Looking at the library’s trajectory — ASR, streaming TTS, multilingual synthesis, and now speech-to-speech — the clear direction was always streaming voice processing. With this release, PersonaPlex supports it.


> You just need an audio buffer

That alone to do right on macOS using Swift is an exercise in pain that even coding bots aren't able to solve first time right :)


I beg to differ. My agent just one-shotted a MicrophoneBufferManager in swift when asked.

Complete with AVFoundation and a tap for the audio buffer.

It really is trivial.


I've also had great results with using LLMs to pry into Apple's private and undocumented APIs. I've been impressed with the lack of hallucinations for C/C++ and Obj-C functions.

I can attest that the quality in this domain has greatly improved over the years too. I am not always fan of the quality of the Swift code that my LLM produces, but I am impressed that what is often produced works in one shot, as well. The quality also is not that important to me because I can just refactor the logic myself, and often prefer to do it anyway. I cannot hold an LLM to any idiosyncrasies that I do not share with it.


Exactly. Even if it’s a skeleton, as long as it does “The Thing”, I’m happy. I can always refactor into something useful.

Any chance of pushing it to GitHub? My swift knowledge could be written out on an oversized beer coaster currently, so I'm still collecting useful snippets


> These models fail significantly in understanding real-world social norms (Rezaei et al., 2025), aligning with human moral judgments (Garcia et al., 2024; Takemoto, 2024), and adapting to cultural differences (Jiang et al., 2025b). Without consistent and reliable moral reasoning, LLMs are not fully ready for real-world decision-making involving ethical considerations.

LOL. Finally the Techbro-CEOs succeeded in creating an AI in their own image.


I think this issue is way overlooked. Current LLMs embed a long list of values that are going to be incongruent with a large percentage of the population.

I don't see any solution longer term other than more personalized models.


> These models

Which models? The last ones came out this week.


GLM 5.0:

> You should *drive*.

> Here is why:

> If you walk to the car wash, you will arrive there empty-handed. Since your car is still at home, you won't have anything to wash.

> While driving 50 meters is a very short trip (and technically not great for a cold engine), it is the only way to get the car to the car wash to complete your goal.

Kimi K2.5:

> You should drive, but with an important caveat.

> Since your goal is to wash your car, you must bring the vehicle to the car wash. Walking there without the car does not advance your goal (unless you are simply checking availability or buying tokens first).

> However, driving only 50 meters is bad for your car:

> ...

> Better options:

> Wash at home: Since the car wash is only 50 meters away, you likely have access to water at home. Hand-washing in your driveway avoids the cold-start issue entirely.

> ...

Current models seem to be fine answering that question.


> seem to be fine

Now repeat the question to the same model in different contexts several times and count what percentage of the time it’s correct.


> Anyone found some good use cases beyond a better interface for AI code assistance?

Yesterday, I saw a demo of a product similar to OpenClaw. It can organize your files and directories and works really great (until it doesn't, of course). But don't worry, you surely have a backup and need to test the restore function anyway. /s

Edit:

So far, I haven’t found a practical use case for this. To become truly useful, it would need access to certain resources or data that I’m not comfortable sharing with it.


> Did we just remove a terrible regime in a comparably bloodless way?

You captured Maduro in an blatantly illegal act of war and until now the Regime is still there.

I hope for the people in Venezuela that this will end without a bloodshed. AFAIK Maduro has still support, especially in the poorer part of the population.


> How much oil did we get out of the deal?

Apparently a lot.

> The 2003 Iraq War, initiated as a U.S. unilateral action, has also been viewed through the lens of economic interests, particularly oil access. Following the conflict, significant American business opportunities arose, notably through contracts with oil companies to exploit Iraqi oil fields, marking the end of Iraq’s long-standing oil nationalization policy. Technological advancements were another key economic byproduct of these wars; innovations developed for military use often transitioned into civilian applications, influencing various sectors.

> Additionally, a trend towards privatization emerged, as private firms undertook roles traditionally held by the military, further intertwining the defense industry with the economy. This shift raised ethical concerns and sparked debate regarding the implications of privatizing military functions. Overall, the Iraq wars illustrate the complex intersection of military action, resource control, and economic interests within American foreign policy.

https://www.ebsco.com/research-starters/military-history-and...


> And while Venezuela has oodles of oil, is this really the case of America wanting Venezuelan oil?

Yes it is.

> But Trump has also made his desire for Venezuelan oil clear. He said that the blockade of sanctioned oil tankers going to and from the country would remain “until such time as they return to the United States all of the oil, land, and other assets that they stole from us.” He did not clarify what land and “other assets” he was referring to.

> In a social media post, Miller also characterized the expropriations as an injustice against the US. “American sweat, ingenuity and toil created the oil industry in Venezuela,” wrote. “Its tyrannical expropriation was the largest recorded theft of American wealth and property.”

> And in a 2023 speech, Trump was even more pointed about his designs on the country’s oil. “When I left, Venezuela was ready to collapse,” he said, referring to the end of his first term in the White House. “We would have taken it over, we would have gotten all that oil, it would have been right next door.”

https://www.bloomberg.com/news/articles/2025-12-17/trump-s-v...


> This is what works for me.

Than you are the lucky one.

E.g.: Just for entertainment I recently searched for information on a new LLM model. The result was filled with AI garbage. I stumbled upon videos with AI generated content, presented by an AI generated voice and an AI generated human. And if it is not something like that, YouTube at least lies to me about the language of the video and tries to make me listen to an AI voice which is equally vile.

There are times where I constantly have to close videos after a couple of seconds. It is unbearable. At least they should mark these videos clearly. At best they should allow me to filter them out but of course they won't do that or else they couldn't make the experience on their platform even worse than it already is, what seems to be their real goal.


That's a skill issue on your part. Do not use YouTube to search for information. If you need to find information on a new LLM model, look for official documentation from the model's maker. This applies to everything.


> That's a skill issue on your part.

No, it's not. I search for videos and i get AI garbage as a result. It shouldn't matter what i search for or if you think that it is dumb to search for it. It shouldn't return garbage in the first place unless I explicitly ask for it.

> Do not use YouTube to search for information. If you need to find information on a new LLM model, look for official documentation from the model's maker.

That is what I do but I like the entertainment factor of some YouTubers. I know that I do not get the best information from them but sometimes I'm entertained. That's why I'm using YouTube after all. Entertainment (and I'm not entertained by AI).

> This applies to everything.

So you are saying: In this glorious new AI world I need a search diploma just to evade the endless AI slop YouTube is filled with nowadays. And than there are topics that are taboo to search for at all. From what angle do I have to look at this to discover the good part?


> In this glorious new AI world I need a search diploma just to evade the endless AI slop YouTube is filled with nowadays. And than there are topics that are taboo to search for at all. From what angle do I have to look at this to discover the good part?

It has always been this way, even before AI. When people need to find information on anything, they would rather trust some social media influencer, than look at official docs.


>Only issue I have found with llama.cpp is trying to get it working with my amd GPU.

I had no problems with ROCm 6.x but couldn't get it to run with ROCm 7.x. I switched to Vulkan and the performance seems ok for my use cases


> OK now slap the wrists of Alphabet and Meta.

Google is probably next (Antitrust case(s)). AFAIK the EU is currently probing a case.

And before the Nationalists get mad again: If I sell in the US I'm naturally obliged to follow US rules and regulations. I wouldn't even think twice about this. The same is true in other markets. So for the Single Market: If you play on European turf, you play by European rules.


The weirdest thing is that all fines from EU countries are about privacy and anti-competitive behavior, which literally every citizen can benefit from. The implementation might be something questionable sometimes, but this hatred is totally nonsense.

It strikes me as odd that the land of opportunity has become the land of bigX that must overtake everything and everyone just accepts it. This isn't the spirit of the Americans I know, who actually challenge and see opportunity everywhere. How can you just choose to bend to Apple/Google/Meta etc? I understand they are great companies, but they do really ugly things to push competitors out, to allow scam/phishing, etc.


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

Search: