HNHacker News
TopNewBestAskShowJobs

dryarzeg

213 karma · joined December 21, 2025

submissionscomments
dryarzeg··on You Do Not, in Fact, Have to Hand It to Them
Yes, we need more children working on factories like in good ol' 1800s

/s

dryarzeg··on You Do Not, in Fact, Have to Hand It to Them
> satanism

I haven't seen this one in the wild (while seeing others) LOL. How does this even work? I mean, like... "if you download the film/LLM you will summon a demon"? I honestly can't understand : )

dryarzeg··on You Do Not, in Fact, Have to Hand It to Them
I agree with you on most parts, but first, not everyone can actually run a model on your own computer that will do 98% of what you need - for most people it's rather around 40% of what you need, depending on the exact model and method of usage; and second, there's still these 2% remaining. And don't forget that not everyone (I mean, not every casual user) can do local inference just due to the lack of technical skills.

And by the way, who will produce this "model that you can run on your own computer"? Although, to be honest, right now in most cases that's not Meta or OpenAI anyway, even though they have some (outdated) open-weight models.

dryarzeg··on European Commission investigating breach after Amazon cloud account hack
Citing from this particular article (emphasis mine):

> While the Commission has yet to share any details about this breach, the threat actor who claimed responsibility for the attack reached out to BleepingComputer earlier this week, stating that they had stolen over 350 GB of data (including multiple databases).

They didn't disclose how they breached the affected accounts, but they provided BleepingComputer with several screenshots as proof that they had access to information belonging to European Commission employees and to an email server used by Commission employees.

The threat actor also told BleepingComputer that they will not attempt to extort the Commission using the allegedly stolen data as leverage, but intend to leak the data online at a later date.

dryarzeg··on TurboQuant: Redefining AI efficiency with extreme compression
If in short, for many inference tasks the bottleneck is memory bandwidth. Suppose you have a machine with a memory bandwidth of 256 GB/s, and let's say you want to do inference for 4B model (model with 4 billion parameters). If you will load the model in BF16 format (16 bits), each forward pass (i.e. each token generated) will require roughly ~8 GB of memory bandwidth. So, 256/8 = 32 t/s, and that's the generation speed you will be strictly capped at even if your processing power is measured in exaFLOPS. But let's say now that you have decided to instead quantize the model and then run the quantized version. Suppose you have made a Q4_K_M version (4 bits + some weights will take more). Now each of your forward passes will take roughly 2-3 GB (rough approximations, reality is different) of memory bandwith (actually, it will be around 2 GB), and even in the worst case 256/3 = 85.3, while 256/2 = 128 t/s. Quants can reduce quality of the model and lower it's performance, but in most modern quantization methods those losses are usually negligible (although, of course, they're still present). So, as you can see, it can be concluded that quantization "widens" (it's not removing it fully) memory bottleneck while still preserving (not always though) acceptable quality.

(Sorry for my terrible English, it's not my native language)

dryarzeg··on FCC has banned the import of all new foreign-made routers here's what you can do
Will he do this for Chinese manufacturers? I'm not sure, but I think they have a decent market share. If this is true, then, I guess, there will still be some missing supply, at least in the short term.

Now, that's not even the main problem. Imagine if someone did something like this with RAM or NAND. Again, maybe I'm stupid, but this sets a precedent. It's not currently possible to move RAM manufacturing (fully satisfying the demand) to the USA, though, so I doubt anyone sane would try to do so.

dryarzeg··on FCC updates covered list to include foreign-made consumer routers
(reposting this comment from the other thread which discusses the same topic, I'm sorry for any trouble that may cause)

From one side, that sure... does have some point. Well, I mean, one could potentially install some kind of a backdoor on the networking hardware they produce, and if it's state-controlled, then it could potentially be a threat.

From the other side, though...

That's crazy. Maybe I'm missing something obvious, or maybe I'm just stupid, I don't know; but at this point, with almost no manufacturing in the USA, this feels like shooting yourself in the foot. Or rather, it's like shooting yourself straight up in the head if manufacturing will not be efficiently (so it can satisfy the demand) moved to the USA (which is a big challenge).

dryarzeg··on The US bans all new foreign-made network routers
From one side, that sure... does have some point. Well, I mean, one could potentially install some kind of a backdoor on the networking hardware they produce, and if it's state-controlled, then it could potentially be a threat.

From the other side, though...

That's crazy. Maybe I'm missing something obvious, or maybe I'm just stupid, I don't know; but at this point, with almost no manufacturing in the USA, this feels like shooting yourself in the foot. Or rather, it's like shooting yourself straight up in the head if manufacturing will not be efficiently (so it can satisfy the demand) moved to the USA (which is a big challenge).

dryarzeg··on Microsoft's "fix" for Windows 11
> Also, who is paying for Windows in 2026?

Buyers of PCs, laptops and other devices with Windows pre-installed. I guess that the cost of the Windows licence is included in the price of each device.

dryarzeg··on Project Nomad – Knowledge That Never Goes Offline
Well yeah, but, hypothetically speaking (and just for the sake of pure curiosity) - what about "no sun" scenario, such as nuclear winter or something similar?
dryarzeg··on Project Nomad – Knowledge That Never Goes Offline
This situation is not unique to Iran.
dryarzeg··on Java is fast, code might not be
That's why I say "in most cases" - so not always, actually. There might be problems with tools, I'm not trying to deny that. And by the way, what if some (or even most) of the users just don't have enough skill to use the tool properly? Again, there could be a problem with tool, yes, but you can't always blame only tools for mistakes users make.
dryarzeg··on Java is fast, code might not be
Maybe I'm stupid, but I never actually understood people who blame programming languages for bugs in software. Because sure, it's good to have guardrails, but in my opinion, if you're writing a program and there's a bug, unless this bug lies somewhere in implementation of compiler/interpreter/etc, you can't blame the tooling, It's you who introduced this bug. It was your mistake.

It's cool when your tooling warns you about potential bugs or mistakes in implementation, but it's still your responsibility to write the correct code. If you pick up a hammer and hit your finger instead of the nail, then in most cases (though not always) it’s your own fault.

dryarzeg··on An update on Steam / GOG changes for OpenTTD
That is true to some extent. However, let me ask you one simple question: how would you try to search for something if you are not aware of it's existence? In other words, how people that are not aware of existence of open-source projects (such as OpenTTD) are supposed to discover them if they're not searching for them on purpose (which is impossible given that they have no clue about their existence)?

Of course there will be some ways like social media or something else. But that question is what seems to worry many people in our case, in my humble opinion. Remember that most of the planet's population is not even aware of existence of open-source projects and open-source concept itself. So how are they supposed to discover it if they don't know about it? When it's present on platforms like Steam and GOG, it helps to spread the word, but when it's not... Well, I guess that seems to be a problem for some people.

dryarzeg··on No, it doesn't cost Anthropic $5k per Claude Code user
How is this related to the inference, may I ask? Except for some very hardware-specific optimizations of model architecture, there's nothing to prevent one to host these models on your own infrastructure. And that's what actually many OpenRouter providers, at least some of which are based in US, are doing. Because most of Chinese models mentioned here are open-weight (except for Qwen who has one proprietary "Max" model), and literally anyone can host them, not just someone from China. So it just doesn't really matter.
dryarzeg··on How to run Qwen 3.5 locally
For me, the main issue with all kinds of recent "advancements" in LLMs is their lack in ability to generalize and extrapolate existing knowledge; they often can be quite weak when it comes to complex associations. Because while many LLMs can demonstrate sufficient theoretical knowledge in maths and physics - and by "sufficient" I mean at least postgraduate level - they often simply fail to apply this knowledge in fields that are closer to real life. At least, that's what I've seen from my experience. They're fine with theory, but once it comes to application, it's all messed up - even in their main "specializations" such as web development or other software-related tasks. And that's... kinda disappointing for me and even makes me a bit sad. We have a powerful tool, but we can't use it's true potential either because we're using it in the wrong way or because it's architecture cannot support this true potential.
dryarzeg··on How to run Qwen 3.5 locally
That's quite a specific task for local models like these though (I mean mujoco), so it might be underrepresented in the training data or RL. I'm not sure if you will be able to see a significant leap in this direction in the next 0.5-2 years, although it's still possible.
dryarzeg··on We Will Not Be Divided
They've already been using Signal - which is "commercial" app, meaning it's not meant to be used like that - for top-secret (or at least highly sensitive) military communications during the military strikes on Yemen. If that was fake, I apologise, I was deceived. I wouldn't be surprised if things turned out that way again, to be honest. That's something to be expected, actually (IMO).
dryarzeg··on I am directing the Department of War to designate Anthropic a supply-chain risk
Ukrainian young (24 y.o.) man here. Living and working in police 30 kilometres away from the actual frontline.

No, thanks, we don't need those "fully automated kill bots". There's absolutely no guarantee that they wouldn't kill the operator (I mean, the one who directs them) or human ally.

We're pretty much fine with drone technology we have.

But for me personally, that's not the most important point. What is more important - and what almost no one in the Western countries seems to realise (no offence, but many of westerners seem to be kind of binary-minded: it's either 0xFFFFFF or 0x000000, no middle ground at all) - is that on the Russian side, soldiers are not "fully automated kill bots" either. Sure, there's a lot of... let's say - war criminals. Yes, for sure. But en masse they are the same young men that you can see on the Ukrainian side. Moreover, many people in Ukraine have relatives in Russia, and there already were the cases where two siblings were in different armies, literally fighting with each other. So in my opinion, "fully automated kill bots" are not an option here. At least unless you deploy them in Moscow and St. Peterburg to neutralize all of the Russian elites, military commandment and other decision-making persons of the current regime.

dryarzeg··on Show HN: Base N Clock - The current time in various number bases
That's really cool, although simple. Thank you very much for sharing that!
dryarzeg··on 0 A.D. Release 28: Boiorix
> It's got robots.

"Shut up and take my money, I'm making a donation" :)

dryarzeg··on Google restricting Google AI Pro/Ultra subscribers for using OpenClaw
I don't see it that frequently. Maybe because I simply do not exist in the places where most of these spambots do their dirty job of killing anything that is alive by the waves of endless nonsensical spam. I agree with you on how... well, I don't even have proper polite words to describe all of that mess. I just wanted to point out that the post actually makes a valid point, though likely it was stolen from somewhere on the Internet.
dryarzeg··on Google restricting Google AI Pro/Ultra subscribers for using OpenClaw
Although I myself am not sure whether this is a real person or a bot, the point seems to be at least somewhat valid to me. I think that some people have become too accustomed to the idea that they can get good things for free or at a reduced price, without thinking about how the economy of production/service they rely on works.
dryarzeg··on Ask HN: Cognitive Offloading to AI
That reminds me of one joke that I heard recently:

- Dad, dad, do you and your colleagues use artificial intelligence at work?

- Oh, dear, we don't even use natural intelligence...

As they say, every joke has a grain of truth in it. There are people who, in principle, do not think about why they do something or what the consequences will be. That's just the way they are.

dryarzeg··on Ask HN: Cognitive Offloading to AI
Good try, tovarish major. Good try.
dryarzeg··on Ask HN: Submission not going live, not flagged
The one that was 3 hours ago is actually marked as [dead]. I don't know the reasons, but obviously it was "killed" for something.
dryarzeg··on Cloudflare outage on February 20, 2026
DaaS - Downtime as a Service©

Just joking, no offence :)

dryarzeg··on I want to wash my car. The car wash is 50 meters away. Should I walk or drive?
Not really, but even if it would be true, I don't think humans ever explained to each other why do you need to drive to car wash even if it's 50 meters away. It's pretty obvious and intuitive.
dryarzeg··on Qwen3.5: Towards Native Multimodal Agents
I'm using Firefox on Linux, and I see the white text on dark background.

> I might have "dark" mode on on Chrome + MacOS.

Probably that's the reason.

dryarzeg··on I’m joining OpenAI
I'm genuinely sorry to bring that up, but you need to work on naming and branding...
← PreviousPage 4 of 5Next →