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

in the PDF metadata we find the proper and appropriate title of this work:

    Nearly SNFS-Speed Signature Forgery Sans Factoring N (NSNFSSSFSFN)

reply


I will admit that when I heard they only had forced blowup I went back to sleep (but I've never been more that cursorily interested in analysis).

reply


I want Opus to be only marginally better, but I do mostly research engineering and its ability to not fuck up my projects is absent. Every time my credits lapse I let kimi and composer2.5 have some play and it’s basically just an excuse for me to keep playing computer because when the oai/ant credits refresh I always need to spend hours recovering from the other models either misconceptions or boneheaded eng practices. Even when I only let it touch my web games…


You have to revert to Opus 4.5 and 4.6. I bet you'll see a massive improvement based on what you're describing


separately trained experts can surpass performance in their activated regime and DOES result in a smarter model, the Claude system cards talk about this and eg there is https://openreview.net/forum?id=iydmH9boLb to read...


There's only so much to do and coordination costs (already burdensome) become overwhelming.


i've been missing this knowledge! thank you for the recommendation.


local laws forbidding facial recognition tech have never been wiser


I was somewhat surprised to learn that phi is _merely_ (1 + √5)/2, I didn't have a good conception of what it was at all but I didn't think it was algebraic.


Phi is conceptually defined like so:

    Suppose you have a rectangle whose side length ratio is ϕ. You draw a line across the rectangle which divides it into a square and another rectangle.

    Then the side length ratio of the new, smaller rectangle is also ϕ.
The diagram is straightforward to set up:

       a        b
    +-----+--------+
    |     |        |
    |  ϕa-|        |
    |     |        |-b
    |     |        |
    +-----+--------+
     \            /
      -----  -----
           \/
           ϕb
This gives us a system of two equations:

    ϕa = b
    ϕb = a + b
If you substitute b = φa into the other one, you get

    ϕ(ϕa) = a + ϕa
And since a is just an arbitrary scaling factor, we have no problem dividing it out:

    ϕ² = 1 + ϕ
Since we defined φ by reference to the length of a line, we know that it is the positive solution to this equation and not the negative solution.

(Side note: there are two styles of lowercase phi, fancy φ and plain ϕ. They have their own Unicode points.

HN's text input panel displays ϕ as fancy and φ as plain. This is reversed in ordinary text display (a published comment, as opposed to a comment you are currently composing). And it's reversed again in the monospace formatting. (Which matches the input display.)

The ordinary text display appears to be incorrect, going by the third usage note at https://en.wiktionary.org/wiki/%CF%95 )


HN's text input panel displays ϕ as fancy and φ as plain. This is reversed in ordinary text display (a published comment, as opposed to a comment you are currently composing). And it's reversed again in the monospace formatting. (Which matches the input display.)

I'm glad you posted this. I'm not a unicode expert and have always assumed these weird dichotomies were some sort of user/configuration error on my part. Realizing the unicode glitches are actaully at the website end instead of between my ears is quite a relief.


To be more specific, that usage note strongly suggests that the problem is in the font used by HN. The font is what complies or doesn't comply with the Unicode standard. We can also say that HN has a problem, but HN's problem is "they're using a noncompliant font for monospaced text".

(On further investigation, I got the characters backwards, and HN's ordinary display is correct while the monospaced display isn't.)


What's stopping you at pasting only a single file? I use the workflow Elon suggests (although I've never used it with Grok) predominately, it's well over 30% of my use of LLMs. I have a small piece of python called "crawlxml" that filters + dumps into <file> tags. And of course the LLM doesn't need your actual code in its context to do its job.


There's no way I'm going to go through my repo dependency tree and paste twenty files into grok one by one.


well, your loss then. clearly your work steps aren’t big enough to benefit from SoA LLM


My work steps are too big to sit around pasting my repo into a text box every time I have a task. This is why integrated IDEs are taking off.


everyone in this thread needs to read this paper: https://dl.acm.org/doi/abs/10.1145/3411497.3420225

Where’s Waldo as presented isn’t even a proof of knowledge


I think the Where's Waldo example, while not technically zero knowledge, gives a pretty good intuition of the idea behind it.

It certainly gives a "layperson" example of being able to prove you know something without revealing it, which isn't the whole definition of ZK but is the idea driving it.


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

Search: