Hacker Newsnew | past | comments | ask | show | jobs | submit | imurray's commentslogin

The 5-digit years are possibly inspired by the Long Now Foundation https://en.wikipedia.org/wiki/Long_now

Tacking on to this, one zero, besides being sufficient to serve the foundation’s purpose and ever so slightly practical, is especially meaningful to humans because it isn’t actually about any longer period of time than we’ve already been at this, while two or more would be. Civilization and the “age of humans” are about 10,000 years old, the biological species 300,000, though we weren’t the dominant hominid or growing exponentially until about 100,000 (99,999) ago.

Because replacing a delimited variable-width encoding with a fixed-width one is supposed to help, somehow...

Here's a more recent implementation of rock paper scissors with an Aaronson Oracle: https://rpsoracle.netlify.app/example/


Ooh, thanks for this. Also, long pressing "My apps", or pressing the (+) next to it, lets us create a homescreen icon directly to "My apps".


The reference implementation of BLAS is in Fortran, but these days fast implementations (using the same interface) tend not to be. OpenBLAS is in C and much of MKL is in C/C++. An individual widely-used package can be rewritten, much like improving a compiler, or giving it a new output, benefits lots of software all at once.

However, OpenBLAS still links to the reference LAPACK in Fortran, because it's mostly made fast by linking it to a good BLAS implementation. It could be rewritten, or converted to C with f2c. Although there's a long tail of useful but often niche Fortran software, some in Fortran 90 that f2c can't convert. Having a fortran compiler is still useful, and covers all of them at once. Potentially that compiler could compile to another language, to cover more and future platforms all at once (like f2c does). It looks like that's what's going on here, compiling fortran to wasm.


> It shouldn't be too hard for the gov to commission a RosettaStone like german learning app ((suboptimal, i know)) that works somewhat well and give it out for free, right?)

Deutsche Welle (DW) is funded by German taxes and has materials for learning German: https://learngerman.dw.com/ as well as regular new materials aimed at learners, such as the news spoken slowly https://learngerman.dw.com/de/langsam-gesprochene-nachrichte... -- although I found it easier to listen to the normal news (e.g. tagesschau) slowed down slightly. (W-h-e-n p-e-o-p-l-e s-p-e-a-k s-l-o-w-l-y i-t c-a-n b-e h-a-r-d-e-r t-o l-i-s-t-e-n t-o!)

Not state-funded, but I recommend https://www.easygerman.org/ -- they have lots of good youtube material, a nice podcast, and an app I've not used. Most of the material is free with bonus materials for supporters.


> although I found it easier to listen to the normal news (e.g. tagesschau)

Tagesschau also has a version in "simple German" and you can turn on subtitles: https://www.tagesschau.de/tagesschau_in_einfacher_sprache


Also promoting German language and culture (I think technically an independant charity but I guess there's a lot of government financial support):

https://en.wikipedia.org/wiki/Goethe-Institut


There's no way to say this without sounding like an old crank; but I think putting down your phone when on the U-Bahn and just reading the news and/or ads on the little in-train screens helps a surprising amount!


John Skilling has used a generalization of the Hilbert curve to n-dimensions in his BayeSys software for Bayesian inference: https://www.inference.org.uk/bayesys/ -- the manual describes how (pdftex will make a nice pdf), with further references, and C code is available.


Here is a pytorch optimizer that can maintain a matrix as orthogonal throughout optimization:

https://github.com/adrianjav/pogo — POGO: A Proximal One-step Geometric Orthoptimizer

https://arxiv.org/abs/2602.14656 — An Embarrassingly Simple Way to Optimize Orthogonal Matrices at Scale; Adrián Javaloy, Antonio Vergari


That's useful, but wouldn't help with this particular experiment because they orthogonalize activations, not weights


I think you're thinking of "You Suck at Excel" by Joel Spolsky. https://www.youtube.com/watch?v=JxBg4sMusIg

Lots of past HN discussion... https://hn.algolia.com/?q=you+suck+at+excel

That video was before the introduction of xlookup, which I believe is now worth considering over index and match.


Yes! Thank you!


That version has security fixes from the same day as the latest rsync release: https://ubuntu.com/security/notices/USN-8283-1

As usual, Ubuntu backported fixes and didn't upgrade to a new version. Whether or not they also backported regressions in edge cases that afflict the latest rsync, I don't know. Pinning the Ubuntu package may prevent getting further regressions, but is preventing you getting any future such backported security fixes.


The Ubuntu backport did include regressions that an Ubuntu update a couple of days later addresses: https://ubuntu.com/security/notices/USN-8349-2


For those wanting alternatives to KL-divergence, the KL and Jensen–Shannon divergences are both F-divergences: https://en.wikipedia.org/wiki/F-divergence


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: