Hacker News: nenaoki

New comment by nenaoki in "Show HN: PinSend – Share text between devices using a PIN(P2P, no login)"

nenaoki — Tue, 03 Jun 2025 19:09:55 +0000

I think they just use the encryption and key exchange that WebRTC has cooked in https://datatracker.ietf.org/doc/html/rfc5764

New comment by nenaoki in "Show HN: PinSend – Share text between devices using a PIN(P2P, no login)"

nenaoki — Tue, 03 Jun 2025 17:34:47 +0000

The pin would just be for coordination, not encryption.

New comment by nenaoki in "Modification of acetaminophen to reduce liver toxicity and enhance drug efficacy"

nenaoki — Sat, 24 May 2025 06:18:44 +0000

it's sad to compare.

apart from that: the word "mark" comes from a root for "boundary" or "border", and really it doesn't need to be about that; we're all in this together.

New comment by nenaoki in "Positional preferences, order effects, prompt sensitivity undermine AI judgments"

nenaoki — Sat, 24 May 2025 05:49:35 +0000

your shrewd idea might make a fine layer back up the Tower of Babel

WaSCR: A WebAssembly Instruction-Timing Side Channel Repairer

nenaoki — Mon, 12 May 2025 21:50:00 +0000

Article URL: https://dl.acm.org/doi/10.1145/3696410.3714693

Comments URL: https://news.ycombinator.com/item?id=43967806

Points: 1

# Comments: 0

New comment by nenaoki in "WASM 2.0"

nenaoki — Sat, 10 May 2025 00:07:26 +0000

The Wasm Constant Time proposal was just moved to inactive 4 days ago[0].

From what I can tell the bulk of the work for it was done in 2018[1], but it needs updating to consider SIMD, and for legwork to be done on moving it along as a proper spec extension.

Until someone picks up this valuable work and lands this much-needed feature in Wasm, we're extremely vulnerable to timing attacks in all Wasm crypto.

[0] https://github.com/WebAssembly/proposals/blob/9fc7a85e/inact...

[1] https://github.com/PLSysSec/ct-wasm

New comment by nenaoki in "My imaginary children aren't using your streaming service"

nenaoki — Sun, 13 Apr 2025 10:00:38 +0000

Python devs needing to be sheltered from mature C++ themes?

Checks out I guess.

New comment by nenaoki in "Qwen2.5-VL-32B: Smarter and Lighter"

nenaoki — Tue, 25 Mar 2025 10:37:59 +0000

Just based on the stage of the game I'd say it's not likely, but the possibilities are there:

https://news.ycombinator.com/item?id=43121383

It would have to be from unsupervised tool usage or accepting backdoored code, not traditional remote execution from merely inferencing the weights.

New comment by nenaoki in "Qwen2.5-VL-32B: Smarter and Lighter"

nenaoki — Tue, 25 Mar 2025 10:22:48 +0000

My understanding is that models are currently undertrained and not very "dense", so Q4 doesn't hurt very much now but it may in future denser models.

New comment by nenaoki in "Carefully but Purposefully Oxidising Ubuntu"

nenaoki — Tue, 18 Mar 2025 04:22:42 +0000

Pretty absurd conflation you've got there.

The package manager just offers a common interface for interop, you can still build without dependencies.

New comment by nenaoki in "Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs [pdf]"

nenaoki — Wed, 26 Feb 2025 05:15:39 +0000

(NB "Two-faced LLMs" are apparently trivial: https://news.ycombinator.com/item?id=43121383)

New comment by nenaoki in "Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs [pdf]"

nenaoki — Wed, 26 Feb 2025 04:48:12 +0000

An LLM would have to be two-faced in a sense to surreptitiously mess with code alone and be normal otherwise.

It's interesting and a headline worthy result, but I think the research they were doing where they accidentally found that line of questioning is slightly more interesting: does an LLM trained on a behavior have self-awareness of that behavior. https://x.com/betleyjan/status/1894481241136607412

New comment by nenaoki in "Intel's Battlemage Architecture"

nenaoki — Wed, 12 Feb 2025 17:38:27 +0000

These are a bit mythical, finding one for sale is no small feat.

I guess adding memory to some cards is a matter of completely reworking the PCB, not just swapping DRAM chips. From what I can find it has been done, both chip swaps and PCB reworks, it's just not easy to buy.

Software support is of course another consideration.

New comment by nenaoki in "Intel's Battlemage Architecture"

nenaoki — Wed, 12 Feb 2025 15:29:18 +0000

>A clamshell 5090 with 2GB modules has a max limit of 64GB

How does "clamshelling" get around the 32-bits per module requirement? Do the two 2GB modules act as one 4GB module when clamshelled?

New comment by nenaoki in "Are LLMs able to notice the “gorilla in the data”?"

nenaoki — Sat, 08 Feb 2025 19:48:05 +0000

Maybe A Whisker Away is actually a documentary...

New comment by nenaoki in "Official DeepSeek R1 Now on Ollama"

nenaoki — Tue, 21 Jan 2025 07:25:37 +0000

https://nvidianews.nvidia.com/news/nvidia-puts-grace-blackwe...

GB10, or DIGITS, is $3,000 for 1 PFLOP (@4-bit) and 128GB unified memory. Storage configurable up to 4TB.

Can be paired to run 405B (4-bit), probably not very fast though (memory bandwidth is slower than a typical GPU's, and is the main bottleneck for LLM inference).

New comment by nenaoki in "IsMyXFeedFucked – Analyze How Your X Feed's Impacting You"

nenaoki — Sun, 19 Jan 2025 14:23:17 +0000

I reckon it's mostly noise. An ad I saw three times in the sample was just an awful mobile game I don't deserve to be subject to ads about, and the other ad I saw twice was just based on my location.

New comment by nenaoki in "IsMyXFeedFucked – Analyze How Your X Feed's Impacting You"

nenaoki — Sun, 19 Jan 2025 11:25:44 +0000

Nice job. It should ignore ads imho, my top 2 "largest influences" were just ads. I mean, I guess we do influence ads via engagement or whatnot, but it's not as if I pay mind to an ad the way I pay mind to a post.

My card: https://files.catbox.moe/t8rzlb.png

New comment by nenaoki in "Kokoro TTS – A lightweight (82M params) text-to-speech model"

nenaoki — Fri, 17 Jan 2025 07:38:56 +0000

https://kokorotts.org/ is the proper site.

New comment by nenaoki in "ASCII porn predates the Internet but it's still everywhere (2019)"

nenaoki — Thu, 02 Jan 2025 12:40:48 +0000

4chan's /jp/ has [sjis] bbcode tags for formatting ASCII art consistently with the Mona typeface.

No thread up though, and the archive doesn't support the tags.