Hacker News: sva_

New comment by sva_ in "I want to live like Costco people"

sva_ — Thu, 07 May 2026 22:17:23 +0000

As a European, I think I mostly liked Costco when I visited. But what I'll always remember is that pizza slice you can get when you leave. The amount of fat and especially salt made me feel like I'm about to have a stroke. I can totally understand how some Americans are unhealthy/obese. It was overall a great experience - 10/10 would do again.

I can't imagine you could buy a pie of that shit to take home.

New comment by sva_ in "Natural Language Autoencoders: Turning Claude's Thoughts into Text"

sva_ — Thu, 07 May 2026 21:31:54 +0000

So the way this works seems to be that you first have an "activation verbalizer" model that generates some tokens describing the activation, and then an "activation reconstructor" that tries to recreate the activation vector. If that reconstruction is close to the original activation vector, they claim, the verbalization probably carries some meaningful information.

I find the fact that this only looks at the activations of some specific layer l a bit interesting. Some layer l might 'think' a certain way about some input, while another later layer might have different 'thoughts' about it. How does the model decide which 'thoughts' to ultimately pay attention to, and prioritize some output token over another?

New comment by sva_ in "ZAYA1-8B matches DeepSeek-R1 on math with less than 1B active parameters"

sva_ — Thu, 07 May 2026 13:11:46 +0000

So at the heart of this architecture is what they call 'Markovian RSA', a combination of two papers RSA[0], which generates a certain amount of reasoning traces for a prompt; and the 'Markovian Thinker'[1] which seems to basically cut the end of those traces to keep context at a reasonable length.

I feel like there's potential to improve that part of just cutting a tunable amount (τ) of tokens off the tail end of those traces, because you may potentially lose valuable insight earlier in the trace? They did train the model (in SFT) to put the relevant information into the tail (τ) of the trace, but I'm not sure this is the best possible way.

0. https://arxiv.org/pdf/2509.26626

1. https://arxiv.org/pdf/2510.06557

New comment by sva_ in "Vibe coding and agentic engineering are getting closer than I'd like"

sva_ — Wed, 06 May 2026 22:07:28 +0000

I wonder if '2000 LOC' was chosen to refer to this old anecdote from the 80s:

https://www.folklore.org/Negative_2000_Lines_Of_Code.html

MV Hondius Hantavirus Outbreak

sva_ — Wed, 06 May 2026 15:36:25 +0000

Article URL: https://en.wikipedia.org/wiki/MV_Hondius_hantavirus_outbreak

Comments URL: https://news.ycombinator.com/item?id=48037462

Points: 2

# Comments: 0

New comment by sva_ in "HERMES.md in commit messages causes requests to route to extra usage billing"

sva_ — Wed, 29 Apr 2026 20:19:24 +0000

Because it hit HN frontpage ...

New comment by sva_ in "Claude.ai unavailable and elevated errors on the API"

sva_ — Tue, 28 Apr 2026 20:40:53 +0000

> prevent Claude Code source code from leaking

That's silly. It's a JavaScript app, they are more or less open source by design. There was no secret sauce in Claude Code.

New comment by sva_ in "TurboQuant: A first-principles walkthrough"

sva_ — Mon, 27 Apr 2026 10:01:01 +0000

Ask Jürgen Schmidhuber

New comment by sva_ in "GPT‑5.5 Bio Bug Bounty"

sva_ — Sat, 25 Apr 2026 15:55:36 +0000

> We will extend invitations to a vetted list of trusted bio red-teamers

Had to chuckle. This sounds like a rather exclusive group?

New comment by sva_ in "New 10 GbE USB adapters are cooler, smaller, cheaper"

sva_ — Sat, 25 Apr 2026 15:21:05 +0000

That's certainly impossible as even USB4 is only 40Gb/s~5GB/s, and of that you could only expect to get 32Gb/s~4GB/s. Or realistically even less due to overhead.

It is probably the speed of it being read into RAM.

Try entering sync right after copying to see how long it really takes

New comment by sva_ in "New 10 GbE USB adapters are cooler, smaller, cheaper"

sva_ — Sat, 25 Apr 2026 10:38:53 +0000

It can draw a lot more under load? https://support.apple.com/en-gb/103253

New comment by sva_ in "New 10 GbE USB adapters are cooler, smaller, cheaper"

sva_ — Sat, 25 Apr 2026 10:34:44 +0000

My laptop has

    $ upower -i $(upower -e | grep BAT)
    [...]
        voltage-min-design:  11.58 V

And I can charge it via USB-C using a 22.5W powerbank @ 12V (HP EliteBook 845 G10.)

I guess that would be out of spec then?

edit: nvm I didn't see the qualifier 'minimum'

New comment by sva_ in "'Hairdryer used to trick weather sensor' to win Polymarket bet"

sva_ — Sat, 25 Apr 2026 10:17:07 +0000

The sensor readings do look suspicious:

april 6 https://www.wunderground.com/history/daily/fr/mauregard/LFPG...

april 15 even more so: https://www.wunderground.com/history/daily/fr/mauregard/LFPG...

New comment by sva_ in "New 10 GbE USB adapters are cooler, smaller, cheaper"

sva_ — Sat, 25 Apr 2026 07:53:51 +0000

Really? I see plenty when I search for 'usb4 nvme enclosure'

New comment by sva_ in "New 10 GbE USB adapters are cooler, smaller, cheaper"

sva_ — Sat, 25 Apr 2026 07:15:37 +0000

It seems like a lot of laptop manufacturers skipped the USB 3.2 Gen2x2 in favor of USB4/TB4.

New comment by sva_ in "We found a stable Firefox identifier linking all your private Tor identities"

sva_ — Wed, 22 Apr 2026 19:12:30 +0000

Does Tor Browser still allow JavaScript by default? Because if you block execution of JavaScript, you won't be affected from what I understand.

New comment by sva_ in "Edit store price tags using Flipper Zero"

sva_ — Tue, 21 Apr 2026 18:50:53 +0000

Not bringing politics into every discussion challenge: impossible

New comment by sva_ in "John Ternus to become Apple CEO"

sva_ — Mon, 20 Apr 2026 20:59:28 +0000

I believe his name is Tim Apple

New comment by sva_ in "Sauna effect on heart rate"

sva_ — Mon, 20 Apr 2026 15:23:25 +0000

It makes your body cool down, which is desirable for sleep.

New comment by sva_ in "Sauna effects on heart rate"

sva_ — Mon, 20 Apr 2026 14:00:07 +0000

> Motivated to understand the immediate physiological response to saunas, we looked at the same-day effects across ~59,000 daily records from 256 users.

Editorialized title is wrong. n=256