4 ms·

Latest AI Code

47 points by albertzeyer 11y ago

yahyaheee 11y ago
Wish this had an app
rayuela 11y ago
Sooooo what is this supposed to be? Vague title and absolutely no description on the site....
- amelius 11y ago
  I guess it is like Hacker News, but for AI code. I like the tagging mechanism, HN could learn from this :) In fact, with all the AI code available, it should be easy to make tagging automatic :)
jimfleming 11y ago
There's also Gitxiv[0] which matches up Arxiv papers with code. TensorTalk seems to have a wider focus (world generation and Twitter bots) while Gitxiv tends to be focused on active ML research. [0] http://gitxiv.com/ http://gitxiv.com/
dang 11y ago
https://news.ycombinator.com/item?id=11033604 https://news.ycombinator.com/item?id=11033604
- rorykoehler 11y ago
  Judging by the 44 points here and the 95 in the link you provided it's obviously a high value post. Some of us didn't see it the first time (myself included) but are grateful for it the second time. Taking it off the front page because it was posted before is bs. It's on the front page because it's valuable to enough people (that's why it has 44 points). I wonder how many people who would find it interesting and haven't seen it before, now won't see it at all because you buried it?
  - dang 11y ago
    That argument would apply to nearly every duplicate. HN's approach is straightforward: we treat an article as a dupe if it has had significant attention in the last year or so. That's frustrating when there's a post you particularly like, but all we can really do is try to be even-handed in general. The limited spots on the front page are the scarcest resource HN has.
    - rorykoehler 11y ago
      Flagging/tagging and merging while keeping it in it's position on the front page would be a much better solution. Most of us don't sit all day on HN and if something gets upvoted to the front page twice it's reasonable to believe that it is there on merit, both times (barring any automated gaming of the system which is another story). Perhaps a review with the aim of considering a less binary solution to this issue is in order?

Latest AI Code

47 points by albertzeyer 11y ago

yahyaheee 11y ago

Wish this had an app

rayuela 11y ago

Sooooo what is this supposed to be? Vague title and absolutely no description on the site....

amelius 11y ago
I guess it is like Hacker News, but for AI code. I like the tagging mechanism, HN could learn from this :) In fact, with all the AI code available, it should be easy to make tagging automatic :)

jimfleming 11y ago

There's also Gitxiv[0] which matches up Arxiv papers with code. TensorTalk seems to have a wider focus (world generation and Twitter bots) while Gitxiv tends to be focused on active ML research. [0] http://gitxiv.com/ http://gitxiv.com/

dang 11y ago

https://news.ycombinator.com/item?id=11033604 https://news.ycombinator.com/item?id=11033604

rorykoehler 11y ago
Judging by the 44 points here and the 95 in the link you provided it's obviously a high value post. Some of us didn't see it the first time (myself included) but are grateful for it the second time. Taking it off the front page because it was posted before is bs. It's on the front page because it's valuable to enough people (that's why it has 44 points). I wonder how many people who would find it interesting and haven't seen it before, now won't see it at all because you buried it?
- dang 11y ago
  That argument would apply to nearly every duplicate. HN's approach is straightforward: we treat an article as a dupe if it has had significant attention in the last year or so. That's frustrating when there's a post you particularly like, but all we can really do is try to be even-handed in general. The limited spots on the front page are the scarcest resource HN has.
  - rorykoehler 11y ago
    Flagging/tagging and merging while keeping it in it's position on the front page would be a much better solution. Most of us don't sit all day on HN and if something gets upvoted to the front page twice it's reasonable to believe that it is there on merit, both times (barring any automated gaming of the system which is another story). Perhaps a review with the aim of considering a less binary solution to this issue is in order?