That Wiki page is woefully incomplete, only covering the digital applications (and these days digital stuff is all done with FETs, not BJTs). Camenzind's Designing Analog Chips has quite a bit more info.

It's hard to control the exact area of the emitter diffusion in a BJT. Think of how an NPN transistor is made on an IC: there's some substrate that the collector gets connected to, then an area is masked off & the unmasked area is doped enough to make it a P-type silicon region. That doping process will have some variation in depth & the masking is never perfect, some will diffuse a bit past the edges of the mask, or not enough will diffuse in. Then another mask layer is applied, and N-dopant is added in the open space to form the emitter. That's also imperfect, on top of an already imperfect base. So the exact gain of transistors isn't well-controlled, even for a very well-controlled collector voltage.

Instead, IC makers can dope multiple smaller emitters in a bigger base region. Smaller doped regions are easier to control (less time for over-diffusion of the dopant) and having a bunch of them has an extra advantage: you can match the ratio of collectors to emitters from one transistor to the ratio of collectors to emitters of another, and variations in the area of any individual emitter will be a much smaller percentage of the total error since the errors of the different emitters will tend to cancel out.

Also, the high-current capability of a BJT is controlled by the emitter length, not the emitter area. Splitting an emitter into long strips can increase how much current a transistor can handle.

Similar things apply to the collector.

For SPICE modeling or building discrete versions of ICs, use multiple transistors in parallel.

Interesting way to draw them, but not particularly weird. Bipolar processes don't really control emitter:collector ratio very well, so gain often varies widely between two transistors even on the same IC. Having extra emitters (or extra collectors) lets the designer match the number of emitters on one transistor to the number of emitters on another, or create ratios between them. That allows for much more accurate gain matching (or gain ratio matching). There are lots of other tricks that can be done in ICs that aren't really available in discrete components. Camenzind's book "Designing Analog Chips" goes over quite a few of them.

Humans are apes, and therefore we're all ape furries. Since catgirls are girls, and girls are human, and humans are all ape furries, catgirls are furries.

Deniers won't care, they'll just say India is part of the conspiracy. As are the scientists using lasers to measure the distance of the moon with retroreflectors the Apollo missions left. As am I for doubting their bullshit. Etc.

And cap insurance profits & executive compensation instead of premiums. A cap on premiums makes insurance non-viable even for a non-profit if the risk is too high, while a cap in profits lets it be valued appropriately. The cap on executive compensation is needed because without that they'd raise premiums excessively & pay themselves the extra instead of accumulating that as company profit for their stock price.

Not much coverage of Vermont's senators? Did you miss all the Bernie Sanders coverage even when he's not running for president?

Better to just use rm -rf ~/*. No need for sudo to destroy the most valuable data (the user's own files).

These days I mostly apply it to passives. Especially small ones. 0201 are a pain to work with, liable to blow away if you look at them wrong. 0402s are better, but still easy to lose. But if I'm trying to keep size & weight down I tend to need the tiny parts, so it's easier to design to use common values as much as possible and spend $30-40 for a reel of 10,000 of them. I doubt I'll ever run out of 10kΩ resistors!

Non-stick pans tend to be made of aluminum (660°C melting point), sometimes alloyed with some copper to improve thermal conductivity. Aluminum-copper alloys tend to melt in the 500-600°C range. Most aluminum alloys melt at a point which an electric stove can easily reach if left on high. The coils can glow cherry-red pretty easily, which is 815-870°C.

DoH & DoT still leak the domain name (and of course IP address) you're connecting to. The domain name leak can be solved by Encrypted Client Hello but that's still a draft and not turned on for many servers.

I'dv deleted the default, it's never come back.

Yes, it's a fancy way to save a tab. I just leave the tab open. Not a feature I want, so not something I want them to waste limited development time on. It'd be nice if it were through the bookmarks interface, so booarks could save state & history the way tabs do, but that's not what's proposed so I'd rather not have this. PWAs are a workaround to make up for the limitations of bookmarks.

PWAs were a feature I marked "want least". I don't like a cluttered home screen, I'd much rather just use bookmarks.

Surface Mount Technology, obviously!

Except Alpine & those based on it, which uses Linux but not GNU libc or GNU coreutils or GNU BASH... Just musl libc & Busybox. I.e. the entire subject of this thread is one of the non-GNU Linuxes.

Yes, I listed sysvinit for that reason. And Musl instead of glibc. GNU is optional in a Linux distro, except for the kernel's use of a GNU license.

Sure, I should have gone further.

Systemd/GNU libc/GNU Coreutils/GNU BASH/Linux/X11//GTK/GNOME
Systemd/GNU libc/GNU Coreutils/GNU BASH/Linux/X11/GTK/LXDE
Systemd/GNU libc/GNU Coreutils/Zsh/Linux/X11/GTK/GNOME
Systemd/GNU libc/GNU Coreutils/Zsh/Linux/X11/GTK/LXDE
SysVInit/musl/Busybox/tcsh/Linux/csh
Systemd/GNU libc/GNU Coreutils/Zsh/Linux/Wayland/QT/KDE Plasma
Systemd/GNU libc/GNU Coreutils/Zsh/Linux/Wayland/QT/LXQT

etc, etc.

There are thousands of combinations of the possible layers needed to make an OS.

Systemd/GNU/Linux/GTK or Systemd/GNU/Linux/QT, really…

It tastes like death and I love it.

It's often used as the name for ammonium chloride on black licorice. The ammonium chloride also makes a great soldering iron tip cleaner!