mbq

9 ms·

31.
by mbq 11y ago
I like the idea to just show the globe (i.e. orthographic projection); it is familiar, interactive, distortions are not deceiving and naturally disappear with zoom.
32.
by mbq 11y ago
Pretty auditable alternative for $35 https://www.tindie.com/products/WaywardGeek/infinite-noise
33.
by mbq 11y ago
The core problem with the "just a neural net" idea is how to tell a robot what it is supposed to do; the whole AI currently relays on some sort of external harness either training the model or pushing the control algorithm into a
34.
by mbq 11y ago
When used as a login shell, shell is expected to source some init scripts to fix PATH, locale and stuff; if a particular distro is not written with alternative shells in mind this will fail, esentially making you use two shells which is c
35.
by mbq 11y ago
Possible next step in this direction is to get rid of both Xen and unikernel by making a specified Java/Erlang VM that only has syscall-calling op-codes for network traffic and virtual disk I/O, while uses hardware virt goodies to
36.
by mbq 11y ago
There is also little known but very cool MooseFS http://www.moosefs.org/
37.
by mbq 11y ago
There are two camps, one for which it is all about points, and second for which it is all about odds. When failing, first one fails at validation, while the second at performance. Obviously both mostly use the same tools and tricks, though
38.
by mbq 12y ago
Shameless plug; I'm a co-author of a method that leverages adding artificial junk features and removing original ones that are likely nonsense to approximate the set of all features that are relevant to the problem (rather than standar
39.
by mbq 12y ago
Instead of Vincenty's, there is a very nice Meeus method for great circle distance, which has virtually identical accuracy and no loop inside; http://www.abecedarical.com/zenosamples/zs_great_circle_rout...
40.
by mbq 12y ago
There is also unison; it is not automatic (custom inotify script is required to mimic Dropbox, and it will force manual intervention upon conflicts) nor stores history, but handles large blobs, many clients and crappy connections quite well
41.
by mbq 12y ago
Even with proper algorithms, per-file encryption leaks quite a lot of meta-data, sizes and usage patterns; also the cloud provider can transparently delete, corrupt and revert files to previous versions.
42.
by mbq 12y ago
For superfluid mechanics (which they seem to be using as a foundation) Donnelly's "Quantized Vortices in Helium II"'s introduction is a pretty good start. The field is quite hermetic, though.
43.
by mbq 12y ago
Meh, people just like simple solutions (or at least those which sound simple).
44.
by mbq 12y ago
It is more-less globally true; locally it depends on the weather itself -- sometimes you can make up to few weeks and be confident about it from ensemble self-consistency (think desert in dry period), sometimes it blows up few hours into fu
45.
by mbq 12y ago
Also this is only about three countries selected a priori, so not a really good sample, even if the number of individuals is huge...
46.
by mbq 12y ago
Not so long ago elevators had drivers...
47.
by mbq 12y ago
I once was on a workshop organised by and for mathematicians which supposed to be about biology. There was a talk where guy was discussing PDE system describing some colony growth; at some point near the beginning he concluded that the resu
48.
by mbq 12y ago
It is sad that substantial part of this lib assumes flat Earth; it will work well on a city scale, but one has to expect spectacular fails on a longer distances and near the poles :/
49.
by mbq 12y ago
Same holds for btrfs.
50.
by mbq 12y ago
Sure, but IMO enough many systems use NTP only in a sntp manner to mention it.
51.
by mbq 12y ago
There is also systemd-timesyncd and Chrony.
52.
by mbq 12y ago
Also MT still has to be seeded, it is not crypto-safe and its quality/computational load ratio is not too good.
53.
by mbq 12y ago
Still no good; you can get spurious correlations between rng outputs on each core. You need something like L'Ecuyer RNG streams.
54.
by mbq 12y ago
This makes me wonder what is the "default" license that applies to all that minified, not labelled anyhow JS code fetched from random CDNs (this especially applies to ADs)... If JS is software, is it even legal to browse the web w
55.
by mbq 12y ago
The problem here is not the for-loop itself but the time used by the R runtime on executing the mapped function multiplied by the number of iterations (this is BTW the main source of advantage for dynamic and GCed but JITed languages like J
56.
by mbq 12y ago
*apply functions are loops underneath -- they only look better and save you time possibly wasted on growing some dynamically sized output structure. The way of solving slow loop in R is to find package which implements it in C/Fortran
57.
by mbq 12y ago
Java still has arrays limited to 2^31-few and no-one cares. And matrix operations make your code closer to the actual math, so less error-prone.
58.
by mbq 12y ago
Yup; but it may be impacted by some privacy regulations in the future.
59.
by mbq 12y ago
Nah; but here is the config with slightly more comments: https://gist.github.com/mbq/7ddbfc2678961e10bb1f
60.
by mbq 12y ago
It really is not; my dovecot, opensmtpd and dkimproxy (needed almost exclusively to ensure delivery to Gmail) config is 45 lines total (with comments).

More ›