Hacker News: k_sze

New comment by k_sze in "Your ePub Is fine"

k_sze — Mon, 15 Jun 2026 03:33:40 +0000

The author says that "In a perfect world, RMSDK would just stop living in the CSS stone-ages or at least provide some kind of error handling instead of dropping the whole book, but I’m not holding my breath."

This is blatantly wrong.

In a perfect world, RMSDK wouldn't exist in the first place and Adobe would have gone bankrupt and become history at least 10 years ago.

New comment by k_sze in "Malicious npm packages detected across Red Hat Cloud Services"

k_sze — Mon, 01 Jun 2026 22:05:27 +0000

Nice work!

Is there a sensible way to add a cooldown to Docker/Podman image pulling?

New comment by k_sze in "I tried to prove I'm not AI. My aunt wasn't convinced"

k_sze — Thu, 26 Mar 2026 07:25:01 +0000

Another shameless plug for my PeerAuth project, which can also tackle this problem. https://ksze.github.io/PeerAuth/

New comment by k_sze in "How Does Microwaving Grapes Create Plumes of Plasma? (2019)"

k_sze — Thu, 27 Nov 2025 06:22:41 +0000

And this is when we realize that the title deserves a [2019] tag.

New comment by k_sze in "Steam Controller"

k_sze — Thu, 13 Nov 2025 08:01:38 +0000

Wrong layout. The Xbox layout is the only correct layout. I’ll die on this hill. :P

New comment by k_sze in "Meta replaces WhatsApp for Windows with web wrapper"

k_sze — Thu, 13 Nov 2025 07:58:39 +0000

Erm, so why wouldn’t I just use WhatsApp as a PWA? I wouldn’t even need to waste the disk space for yet another browser runtime.

New comment by k_sze in "Users Stuck in YubiKey Re-Enrollment Loop on X (Twitter)"

k_sze — Wed, 12 Nov 2025 20:45:12 +0000

I was very confused, too. I mean, I do have a YubiKey but Twitter/X has never (or for a very very very long time not) asked me to use my YubiKey to authenticate. As far as I could remember, I only need to use OTP from Authy as 2FA to login. So the whole thing smelled really fishy.

New comment by k_sze in "DeepSeek OCR"

k_sze — Mon, 20 Oct 2025 07:10:33 +0000

It's interesting how they use "Gundam" in their variant names. I gather that Gundam-M and Gundam are their most powerful ones.

LongCat-Flash-Thinking, LLM from Meituan (China's Equivalent of Uber Eats)

k_sze — Tue, 23 Sep 2025 08:09:24 +0000

Article URL: https://github.com/meituan-longcat/LongCat-Flash-Thinking

Comments URL: https://news.ycombinator.com/item?id=45344180

Points: 2

# Comments: 1

DeepSeek v3.1 released – single model, thinking and non-thinking modes

k_sze — Thu, 21 Aug 2025 06:41:35 +0000

Article URL: https://twitter.com/deepseek_ai/status/1958417062008918312

Comments URL: https://news.ycombinator.com/item?id=44969767

Points: 26

# Comments: 0

New comment by k_sze in "Japanese scientists develop artificial blood compatible with all blood types"

k_sze — Tue, 03 Jun 2025 02:30:07 +0000

It's odd. It seems like this is not the first Japanese team to have developped artificial blood. I did a quick search and it seems there was another team at least as early as 2019 (https://web.archive.org/web/20201111233217/http://www.asahi....)

So what's different this time?

(Upon further examination, the 2019 team at the National Defense Medical College also had Dr Hiromi Sakai. So why is this news now?)

New comment by k_sze in "Apple M3 Ultra"

k_sze — Fri, 07 Mar 2025 06:16:57 +0000

Why did Apple call it M3 Ultra when it's supposed to be more performant than the current top of the line M4 Max? Why not just call it M4 Ultra?

New comment by k_sze in "QwQ-32B: Embracing the Power of Reinforcement Learning"

k_sze — Fri, 07 Mar 2025 02:39:35 +0000

Oddly, the Chinese LLM host SiliconFlow only makes it available with 32k context, which is even smaller than their DeepSeek-R1 offering.

New comment by k_sze in "Kaspersky exposes hidden malware on GitHub stealing personal data"

k_sze — Tue, 04 Mar 2025 06:04:27 +0000

You're confusing Kaspersky with Kasparov, perhaps?

New comment by k_sze in "Kaspersky exposes hidden malware on GitHub stealing personal data"

k_sze — Mon, 03 Mar 2025 04:00:49 +0000

I'm in Canada and I just get auto-redirected to the kaspersky.ca home page when I try to visit the link.

New comment by k_sze in "Surgery implants tooth material in eye as scaffolding for lens"

k_sze — Sat, 01 Mar 2025 07:11:57 +0000

One thing I don't understand is how the surgeon ensures the lens is positioned correctly. We're talking about a rigid lens that can no longer be controlled via muscles, right? Doesn't any misalignment mean you get a fuzzy image at best? And even if you can get a sharp image for some given, fixed distance, you still can't control whether to focus on near things or far things when you look, right??

3FS – a parallel file system from DeepSeek

k_sze — Fri, 28 Feb 2025 01:41:12 +0000

Article URL: https://twitter.com/deepseek_ai/status/1895279409185390655

Comments URL: https://news.ycombinator.com/item?id=43200679

Points: 18

# Comments: 0

New comment by k_sze in "DeepSeek open source DeepEP – library for MoE training and Inference"

k_sze — Tue, 25 Feb 2025 07:11:48 +0000

Practically speaking, is it possible for NVIDIA to "pull the rug" later, intentionally or otherwise, by subtly changing the behaviour of this out-of-doc instruction on new architectures?

New comment by k_sze in "PeerAuth, TOTP-based peer authentication in the post-truth world"

k_sze — Thu, 20 Feb 2025 21:09:12 +0000

I think that the scenario you describe requires one of two conditions:

  1. The attacker knows the time and medium through which the two persons call each other, and have control over the medium, being able to inject themselves;
  2. The attacker coerces one of the two persons to perform authentication.

You have a much bigger problem if any of the above is true.

New comment by k_sze in "PeerAuth, TOTP-based peer authentication in the post-truth world"

k_sze — Tue, 18 Feb 2025 05:32:47 +0000

That sounds interesting! I'm looking forward to seeing what you build.