Josh Horowitz

@qualmist

he/him/his. a.k.a. Mr Regal Jelly. I work at , a simple, innocuous tuple-processing collective.

Vrijeme pridruživanja: ožujak 2015.

Medijski sadržaj

  1. 19. sij
    Odgovor korisniku/ci
  2. 18. sij
  3. 5. sij
    Odgovor korisniku/ci

    That sign has a cameo in the Semi-Charmed Life music video!

  4. 31. pro 2019.

    All of the online SF tree maps are slightly broken? So I made a new one: .

  5. 30. pro 2019.

    One bit I learned from that doc: 20 years before designing the Golden Gate Bridge (sorta), Joseph Strauss made the Aeroscope, a 235ft-high viewing platform for the Exposition which held 120 people as it spun and levered.

    Prikaži ovu nit
  6. 30. pro 2019.

    It is mind-bending that the Marina District used to look like this. Here's a bunch of narrated footage from the Panama-Pacific International Exposition:

    Prikaži ovu nit
  7. 25. pro 2019.
    Odgovor korisniku/ci

    Extremely relatable:

  8. 17. pro 2019.

    The two most poetic foams.

  9. 7. pro 2019.
  10. 6. pro 2019.

    there's an extremely sad row in this spreadsheet; can you spot it

  11. 3. pro 2019.
    Odgovor korisnicima

    I bet that's a lot of us! my guess: going from 0 to 1 is much harder than going from 1 to 2, 2 to 3, etc., so the distribution is bimodal, at 0 and... something else. see attached Mathematica model as crude illustration. (can't find any actual data tho.)

  12. 23. stu 2019.
    Odgovor korisniku/ci

    The bit screenshotted below sounds a lot like the conventional / "textbook" explanation you describe as flawed earlier in the notes. Am I missing a distinction between them?

  13. 13. stu 2019.

    This makes no sense at all, right?

  14. 13. stu 2019.
    Odgovor korisnicima

    Take the histogram of distances of pixels from the origin [A]. Replace each bar with a little triangular bump [B]. Add them up. (This is secretly a convolution.) You get a really good approximation for the data [C]!

  15. 13. stu 2019.
    Odgovor korisnicima

    Here's a rougher dataset extending up to a radius of 10 pixels. The top line connects the integer-radius points & the bottom line connects the integer+0.5-radius points.

  16. 13. stu 2019.
    Odgovor korisniku/ci

    I was really intrigued by your claim that "the circles seem to look nicer when the radius is integer+0.5 rather than an integer". As a rough measure of "niceness", I computed the area of the symmetric difference (XOR) between the pixelated circle and the real circle. Chart:

  17. 3. stu 2019.
    Odgovor korisnicima i sljedećem broju korisnika:

    But you're not the only one who mixed them up...

  18. 2. stu 2019.
  19. 1. stu 2019.
    Odgovor korisniku/ci
  20. 31. lis 2019.

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.

    Možda bi vam se svidjelo i ovo:

    ·