ThouYS

9 ms·

1. 121.▲
   by ThouYS 2y ago

   I'm not an astronomer either, but pretty sure if I generated uniformly random points on the scale of number of visible galaxys, I could find a circle in there
2. 122.▲
   by ThouYS 2y ago

   I've been wondering the same about arms exports for a while. Why wouldn't exported arms all have a killswitch?
3. 123.▲
   by ThouYS 2y ago

   one of the most magnificent books ever written
4. 124.▲
   by ThouYS 2y ago

   That's true. I feel like the important bit is covering the essentials. 2x more is not 2x better, but you need to get above a certain threshold. Incidentally this also holds for some engineering metrics. Some error metrics stop making s
5. 125.▲
   by ThouYS 2y ago

   whoo! that you could jump back up and use variables from below has been an annoyance of mine
6. 126.▲
   by ThouYS 2y ago

   wow. this is it!
7. 127.▲
   by ThouYS 2y ago

   yes, same issue in all transformer tutorials
8. 128.▲
   Garbage collection in Python: things you need to know(rushter.com)

   1 points by ThouYS 3y ago|0 comments
9. 129.▲
   by ThouYS 3y ago

   they don't need to be. dependents will continue functioning, because ABI hasn't changed
10. 130.▲
    by ThouYS 3y ago

    under what circumstances does (2) hold? for vanilla methods it's no problem
11. 131.▲
    by ThouYS 3y ago

    grep truly is the king of tools
12. 132.▲
    by ThouYS 3y ago

    these zero days are already the status quo?!
13. 133.▲
    by ThouYS 3y ago

    This is about synthetic training data generation as much as it is about films
14. 134.▲
    by ThouYS 3y ago

    cool stuff!
15. 135.▲
    by ThouYS 3y ago

    yawn, imagine hating layers of abstraction. lol, lmao even
16. 136.▲
    by ThouYS 3y ago

    "written in node.js" - insert squidward packing his chair meme
17. 137.▲
    by ThouYS 3y ago

    hahaha, can't say I fee sorry for their customers
18. 138.▲
    by ThouYS 3y ago

    this, tags + dynamic views on those tags
19. 139.▲
    Generalized Means for Combining Metrics(thousandyardstare.de)

    1 points by ThouYS 3y ago|0 comments
20. 140.▲
    by ThouYS 3y ago

    onnx is nice in principle, but pretty limited. core ops like non-max-suppression can't get properly converted. also model deployment is not great, memory consumption and control thereof worse than with tensorflow.
21. 141.▲
    by ThouYS 3y ago

    Instantly reminds me of "La Horde du Contrevent" a magical adventure on a planet with constant wind, and wind-scholars
22. 142.▲
    by ThouYS 3y ago

    either I'm getting old, or there is a lot of energy being spent on quasi-useless "improvements" these days
23. 143.▲
    by ThouYS 3y ago

    hahahahahahaha
24. 144.▲
    by ThouYS 3y ago

    good to hear, will this fix the de-facto unusability on macOS? I've used Inkscape since around 2008 and absolutely love it, but had to switch to affinity designer, because the gui is too slow to work with :(
25. 145.▲
    by ThouYS 4y ago

    everything. why are there still cobol programmers? why is c++ still the defacto native language (also in research)? but also I don't see any problem there, I think the python + c++/rust idiom is actually pretty nice. I have a bill
26. 146.▲
    by ThouYS 4y ago

    yawn, it's all rubbish
27. 147.▲
    by ThouYS 4y ago

    Probably McElreath's Statistical Rethinking, or Stanfords famous Neural Net course on youtube
28. 148.▲
    by ThouYS 4y ago

    sounds like henlo to me: https://henlo.app/
29. 149.▲
    by ThouYS 4y ago

    thanks for that link. that pretty much is exactly how I imagine him. I remember reading Ashley Vance's biography. At some point Elon makes a calculation about how many hours a girlfriend would need.
30. 150.▲
    by ThouYS 4y ago

    amazing! the author displays great pedagogical skill to his student