> " could it somehow get access to the subtitles and then use them to answer queries?"
It's not even necessary - computers are already excellent at understanding spoken words. Have you tried automatic captioning recently? Half the inputs to my phone are already voice, not text.
Video is a harder problem, but it's not too far behind.
Regarding in-person I9 checks, I had a very interesting experience recently with a friend who was hired at a big 4 accounting firm. They deputize a friend of the employee (me) to verify that they have checked the I9 in person. They use a service for this: https://www.lawlogix.com/
It was a very strange experience. Here I am a totally unrelated party becoming an "authorized representative" of the firm to look at a document and verify it's real. There was no real verification of me, mind you. I could have been anyone.
I have no idea how normalized this sort of thing has become, but I was quite surprised that they appoint some random individual rather than have an actual employee do the document verification.
Hey, look. As a systems guy myself I understand the aesthetic around clean and simple software. That's a totally reasonable preference.
Here's the thing: GNU userland isn't some poorly designed aberration. It's by far the most popular unix userland in the world and is very commonly ported to systems which may do things differently. Almost every other unix has a distro to supply GNU userland tools on it -- no other unix userland can make the same claim.
Not everyone wants it, sure. Especially people with perspectives broader than the typical user. Of course. But it's the most popular system in the world -- not some weird never-used prototype.
And in the case of yes(1), I think you would have to try very hard to find a reason for a user of the tool to have a negative preference.
Because the functionality in this case is a strict superset, regardless of what you're accustomed to.
I didn't grow up on GNU either, but the GNU yes does everything I expect having used other unix systems.
> This isn't something you can quantify and say 'strictly superior' it's all just taste and opinions.
For a complicated tool, sure. But this tool is very simple. The GNU yes(1) accepts 100% of inputs that work on the BSD tool. It does a few additional things too.
I can't speak to m1 in particular, but BSD grep is known to be very slow regardless of platform. You can just brew install grep to get a fast one, but you'll still always have to always deal with avoiding the slow system grep.
That's a developer-perspective issue. You're talking to a user here, and from a user perspective, the GNU utilities are generally strictly superior.
Take the example you provided, of yes(1). Most of the extra code relates to optimizations, and ensuring that the entire argv is echoed back rather than just the first argument. Because of this, even a very simple example like:
"yes hello world"
will behave differently on BSD vs GNU, with the GNU version likely being what users expect.
I've been playing around with the edge TPU on a coral.ai board. It's very impressive - I'm able to do real time object detection at tens of frames per second vs one frame every 15 or so seconds on my CPU.
The control knobs are extremely easy to clean: I just pull them off and throw them in the dishwasher.
The display is never clean. It has grease on it and I can't ever scrub it hard or use caustics for fear of damaging it.