pastille
@pastille

Not everything has been documented, as the filenames are encrypted. Some significant games known include almost every Angry Birds game, Bloons Tower Defense 5, Club Penguin, and Roblox.

Archive.org links to the leaked files are here:
https://web.archive.org/web/*/https://d193ln56du8muy.cloudfront.net/ipas/*
https://web.archive.org/web/*/https://d3qktfj96j46kx.cloudfront.net/*
https://web.archive.org/web/*/https://builds.testflightapp.com.s3.amazonaws.com/*
You can download them for yourself to help document which files contain which apps. You will need a jailbroken iPhone/iPad on iOS 10 or lower to install and play them, as they are mostly 32-bit apps.

Important edit now that the story has progressed:

As the post is now very outdated I figure I should update it to prevent spreading any misinformation. First, all the files on archive.org have been removed, due to a DMCA request from Apple themselves. Do not try to reupload any files to archive.org. Second as the Angry Birds community doesn't wish to be the center for this information, a standalone server was created. You can find it here: https://discord.com/invite/NMGYQy3teZ

Most interesting findings have been backed up to torrents and Google Drive folders which can be found there.


xorgol
@xorgol

Yet another example of piracy being the best form of archival



belarius
@belarius

Printed on paper nearly 7 meters long (and only 19cm tall), Metamorphosis III is the largest print ever produced by M.C. Escher, undertaken from 1967 to 1968. This also marks the first day of a month honoring Janus, a god of beginnings, endings, and the transitions between them. I hope everyone enjoys this Feast For Janus!

The full print can be viewed here.


anfael
@anfael

in the deepest darkest 90s one of the maths classrooms at my high school had a reproduction of this, probably on just a ream of A4s, taped up around the walls and my eyes would just scan over it back and forth the whole time the teacher went on about First Outer Inner Last and completing the square or whatever




kobi-lacroix
@kobi-lacroix

The paradox of the Dunning-Kruger effect is that it probably doesn’t exist, but those who know that it probably doesn’t exist don’t bring it up as often as those who are certain that it does, thereby proving that it does exist.


belarius
@belarius

Authors Jansen, Rafferty, and Griffiths (2021) have done one of the better recent re-examinations of this now-tired meme, and I wish every person I saw on the Internet name-dropping Dunning-Kruger (DK) as a way of dunking on people they dislike would consider its conclusions seriously.

Conclusion 1: The vast majority of the effect attributed to DK is due to a statistical artifact called regression to the mean (RttM). Simply put, most tests of ability aren't great at measuring that ability, including our own internal self-assessments. When measurement error is applied to both of the scores obtained from your participants, then taking the difference between those scores gives you what looks like DK for free. Put another way, the famous skill-confidence disconnect is almost entirely consistent with nothing more than noisy measures of skill and of confidence. So, big-picture, DK as most people cite it does not exist, because it's a statistical mirage that Dunning and Kruger should have known better than to be fooled by. This is further exacerbated by the quartile maneuver D&K used in their original paper, if anyone's keeping score. Viewed in the cold light of day, the original paper's a bit of a statistical trainwreck.

Conclusion 2: Based on new data using a much larger sample of participants, this new trio of authors show an interesting wrinkle, depicted in my reproduction of their figure, above. On the one hand, their best estimate of the RttM effect mostly describes the data, but another model that incorporates an interaction with performance does a slightly better job. This suggests that, at least on average, people with the very highest and very lowest scores tend to be a tiny bit overconfident, whereas those with average-to-slightly-below-average scores actually tend to be slightly underconfident. Of course, across all these points, there's enormous individual variability.

Conclusion 3: Because confidence varies dramatically from person to person, this very small over-to-under-to-overconfident effect barely registers, and wouldn't help you very much in trying to predict someone's skill level from their confidence (or vice versa). Formally, the "effect size" of this underlying effect is quite small. As such, while there is now some evidence of a small (and more complicated) skill-confidence disconnect, that effect is so much smaller than how much people just vary in their confidence in general that it's probably best to proceed through life assuming that DK-like effects don't exist in a way that should impact your judgments of people.


NireBryce
@NireBryce

it tested Cornell undergrads who were most likely to show up to a study that gave extra credit.

meaning that on average:

  • being an ivy league school, your lower performing students are, in general, taught to think higher than themselves, whereas the underconfident achievers may feel the test is too easy and they're missing something
  • due to the reward being extra credit, you're more likely to attract people struggling with school on one end, and overachievers covering their bases on the other

it's uniquely unsuited for testing what they're trying to test

and that's before we talk about the whole paper sorta feeling like a victory lap of "we set out to confirm our biases and were incredibly successful"