That's a beautiful little project.

We've gone from "So clear you can hear a pin drop" to "Can you hear me now?!?"

Right. Cellular telephony just barely works now. There's lag as long as a second, even when the call supposedly isn't going over VoIP. (Sprint seems to have that problem.) There's occasional echo when the lag exceeds what the echo suppressors can handle. Background noise kills the cellular compression algorithm.

Why don't we have CD-quality audio on phones?

But this is one of the few times when the article is actually more interesting than the comments here. Worth reading, even if that is breaking the unwritten rules around here.

What does Yahoo still do, anyway?

Apple needs to get their ruggedness act together. Meanwhile, here's a real phone, the Caterpillar B15.

Cat B15 tested by users. Dragged behind car. Used to play basketball. (As the ball, not as a computer game.) Dropped off bridge. Run through cement mixer. Frozen in bucket of ice. Run over by car. No problem.

Cat B15 tested by Caterpillar. Dropped into pool of water. Scooped out with heavy equipment. Run over by front end loader. (One of Cat's smaller front end loaders.) No problem.

It's an Android phone. The B15 runs Android 4.2; the new B15Q runs Android 4.4. Price around $300. Available in the US at Home Depot. Unlocked; pick any GSM carrier. T-Mobile works. No annoying carrier-provided apps. Caterpillar preloads apps for ordering Caterpillar heavy equipment parts and renting heavy equipment.

If you have one of these in a pocket, you will break before it will. I carry one of these horseback riding.

I know lots of people who go, but have no desire to go myself.

You say you're an experienced embedded-systems developer. Those are rare. Stay with that and get better at it. There are already a huge number of people grinding out appcrap, more than the app market can support. Soon there will be a glut of former phone app programmers, if there isn't already.

Try to get in on the back end of the "Internet of things". That crowd is overrun with appcrap people and has no clue about embedded.

The low-end 3D printers, the ones that try to weld ABS string together, still suck. TechShop has several of them. The Jet was a a flat failure. The Replicator 2 is OK if you're not building something more than about 2cm thick. I haven't tried the Type A Machines unit. In the end, it's a slow way to make prototype plastic parts that are inferior to injection-moulded ABS. Injection moulding requires machining a die, which is a big job, but then the production rate is high and the cost is very low.

The higher end printers have much better quality and more material options, but the machine cost is high and the process is slow. The really high end printers, the ones Space-X and Lockheed use to print aerospace parts, are very impressive, but still slow.

The real breakthrough in LED lighting is getting rid of electrolytic capacitors in the power supply. Those are currently the components with the shortest life. See "Elimination of an Electrolytic Capacitor in AC/DC Light-Emitting Diode (LED) Driver With High Input Power Factor and Constant Output Current" Variations on that technology are now going into production LED lighting units. This should push unit lifetimes up from 20,000 hours to that of the LEDs, 40,000 or so. (Provided the quality of the LEDs doesn't slip.)

That is right. I was in during the middle 80s when it existed, but was still rare.

It's time to demand that Fourth Amendment rights be taken as seriously as Second Amendment rights. That's starting to happen.

What, they're going to ship a VR headset without positional tracking? When you turn your head, nothing happens? That's not VR. That's a TV you wear on your head.

That's neat. The demo takes in the video from a video game of the Pong/Donkey Kong era, can operate the controls, and in addition has the score info. It then learns to play the game. How to do that?

It's been done before, but not this generally. "Pengi", circa 1990, played Pengo using only visual input from the screen. It had hand-written heuristics, but only needed vision input from the game. So we have a starting point.

The first problem is feature extraction from vision. What do you want to take from the image of the game that you can feed into an optimizer? Motion and change, mostly. Something like an MPEG encoder, which breaks an image into moving blocks and tracks their motion, would be needed. I doubt they're doing that with a neural net.

Now you have a large number of time-varying scalar values, which is what's needed to feed a neural net. The first thing to learn is how the controls affect the state of the game. Then, how the state of the game affects the score.

I wonder how fast this thing learns, and how many tries it needs.

No, it was called a Senior Airman. Very few Sr.Amn went to NCO school and took the rank of buck sgt. before just going for staff, so the majority of E4s were Airman tier, not NCO tier. Sr. Amn and Buck paid the same, both being E4, with the only difference being NCO school and tier, plus some advantages (priority) if you wanted to separate rats and quarters and were single. Oh, and the star on your sleeve was silver instead of blue, although that isn't a perk, just an indication you were an NCO.

Note that most of the "exceptional" airmen back then would simply go below the zone (make E5/Staff Sgt. in less than 4 years) rather than seek buck sgt. Bucks were fairly rare for a variety of reasons, including the above.

Yes, I was in the Air Force. So was most of my family.

FastCGI implementations are supposed to execute the specified executable without any parameters from the HTTP request. The FCGI program then reads and processes multiple HTTP requests, with no shell involvement. Unless the program invoked by FCGI itself invokes the shell (which PHP scripts can do), there should be no problem. I'm not a PHP user; someone with PHP internals expertise needs to look at that world for vunerabilities. Can arguments from the HTTP request make it into the environment of subshells invoked by PHP?