This is an important new thing. We've had question-answering programs working against specific data sets since Bobrow's "Baseball" program of the 1960s. We've had a whole range of question-answering specialist systems running in tandem since Yahoo introduced vertical search around 2005. But cross-topic generality has been elusive.

If this is real, it's a major development. Is there anything better than the Tired article available?

OK, good. Now if ICANN applied that tougher standard to domain name registrars, we'd make progress. But no, ICANN still allows registrars to register domain names without forcing them to comply with the most restrictive profile.

Compare Space-X's launch manifest from a year ago with their current launch manifest. They're six months to a year behind their launch schedule. There were supposed to be three Space-X ISS resupply flights this year, #4, #5, and #6. Flight #4 is currently scheduled for September. There are five commercial customers waiting for their scheduled 2014 launches.

Some of this isn't Space-X's fault, and some of it is. All these are Falcon-9 launches, some with the Dragon capsule. No major new hardware is involved. It's not clear where the holdup is coming from. There have been problems with scheduling at Canaveral. 2014 was supposed to be the year that Space-X caught up on their launch manifest, but that's not happening. Unclear why.

If they had a Microsoft library not authorized for free distribution in their program, Microsoft would be demanding substantial damages.

The web site reads like they're a big consultancy, another McKinsey. Then the testimonals are all about Walter. Oracle manager: "Walter showed a great depth of knowledge in Word, WordBasic Macro programming". He still has recommendations up which mention Turbo Pascal. Not seeing rocket science here. The biggest success reported was translating some large English-only application into multiple languages, which made it valuable in Asia. That's nice, but a routine job. He claims to have written a general-purpose program to help with such jobs.

He also claims to have written ScenGen, a "scenario generator". It looks like that originated at Boeing in the mid-1980s. Running on a Compaq PC with 2MB back then. The pitch for the current model sounds like the one from back then, although the graphics are probably better now.

The web site is awful. There are lines of text with excess white space in the middle. I looked at the HTML, expecting to find some overly complex Javascript which was misbehaving. No. The HTML source just has explicit non-breaking spaces in the wrong places.

He seems to speak at a lot of strange conferences, such as the Family Office Association. A "family office", in this context,is a staff which manages the family fortune for a large, wealthy family. The Rockefellers have one.

This is getting weird.

From low orbit, about 25cm is reported for military satellites. Maybe a little better. DigitalGlobe is now at 41cm. Reading newspaper headlines from orbit is unlikely. If the military satellites were doing that well, there would be little reason to fly recon drones or aircraft.

Once you can recognize vehicles, weapons, and troops from orbit, more resolution doesn't help much militarily. The next step, which is where DigitalGlobe is going, is more frequent imagery, and wider fields of view and more downlink bandwidth without giving up resolution. Digital Globe says they collect 3 million km^2 of imagery per day. That's only 0.6% of the earth's surface.

It's an aimbot for real rifles. Now, any rifleman can be a sniper.

Yes, it's too big, too complicated, and too expensive. That's a temporary problem. Ever see the first laser sight, from the 1980s? It used a helium-neon laser tube and required a power cord. There's been some progress since then. This aimbot technology should be down to smartphone size, if not cost, soon enough.

No, it doesn't "roll all languages into one". It just allows embedding of the text of another language, such as HTML, into a Wyvern program. Variables can be substituted. Like this:

let webpage : HTML = <html><body><h1>Results for {keyword}</h1
<ul id="results">{to_list_items(query(db,
SELECT title, snippet FROM products WHERE {keyword} in title))}
</ul></body></html>

(except that the last 3 lines above should be indented, because this language uses Python-style block notation.)

Of course, everybody does that now, but the way they do it, especially in PHP, tends to lead to problems such as SQL injection attacks. The idea here is that Wyvern has modules for the inserted text which understand what kinds of quoting or escaping are required for the embedded language text.

I just glanced at the paper, but that seems to be the big new feature.

Utilities can only delay solar a little. PV solar, without subsidies, is just now becoming cheaper than fuel-powered electricity in sunny locations. Bloomberg reports the first non-subsidized solar plant to be built in Spain.

In the next decade, we'll see the end of subsidies and continued growth in PV solar. Anywhere the biggest daytime power load is from air conditioning, solar will win out.

So far, all they've found it doing is reporting the IMEI by sending an HTTP GET http://api.account.xiaomi.com/pass/v3/user@id?type=MXPH&externalId=01, The data is transmitted as a cookie of the form deviceId=IMEI . (The API returns a brief reply in JSON.) That tells them the phone has connected to the phone network, and its IP address. That's not particularly interesting information. The carrier knows the IMEI number, too, of course. Perhaps this is to check up on whether carrier-reported sales data matches actual phones coming on the air.

Carriers, app vendors, Microsoft, Google, and Apple collect far more data than that. There are way too many things phoning home with the user's contact list and other personal info.

Right. With mid-tier open source projects, there's a good chance they're either unfinished or abandonware. (Lower-tier open source projects are both.) There's only so much attention available.

It's quite possible to build a flying car. It won't be cost-effective to build or operate, because it will need bizjet-sized jet engines for VTOL. Elon Musk once remarked that he'd like to build one "just for fun". I wish someone would, just to shut everybody up. Quadrotors work just fine, after all. Scaled Composites could probably have something flying in a year. Probably not much range, but flying.

Just because Moller has been failing at this for 40 years doesn't mean it's impossible. That's a problem with Moller.

Unless they committed to that contractually, they might.

That's right. Automatic lane keeping plus radar-based cruise control is right in the middle of the "deadly valley" - good enough to allow hands-off driving 98% of the time, not good enough to handle trouble. This is why that Cruise startup building a budget self-driving system worries me. Thos guys are from "social" apps. They're thinking they can ship something that sort of works, and that's good enough. It isn't.

Auto manufacturers are held to a much higher standard than the computer industry is used to. GM is being sued because their ignition switches could turn off if people hung too much crap on their keychain. (Something unlikely to be caught in testing, because, at the test track, each key hangs on a separate key tag.) "Speeding, cellphone texting, intoxication... irrelevant. We are not looking at the driver, or the circumstances of the driver's negligence. We are looking at the automobile, and only the automobile." - terms of the GM settlement.

The minimum safe level of performance for a self-driving car is that the vehicle must be able to bring itself to a safe stop, preferably at the side of the road, in any emerging bad situation. Even after any single-point failure.

Few computer based consumer products meet that standard, but a some do. The Segway is a good example. There's enough redundancy in a Segway to keep single failures from face-planting the user. Five rate gyros instead of three, two batteries, two processors, and a safety shutdown mode that brings the vehicle to a stop and sounds alarms to tell you to get off before it fails.

Low-level programming is a specialist issue. Maybe it's time to turn C programming over to people with real EE degrees, or who can at least use an oscilloscope and wire up an Arduino. At the application level, who has time to manage memory by hand any more? EEs and mechatronics people, and OS and compiler developers, need to learn C, but most application programmers today do not.

The emphasis on Java isn't unreasonable. The pure-interpreter languages (Python, Perl) are too slow for large server-side operations. (If it's 3x as slow, you may need 3x as many server racks. That costs.) Java is memory-safe and goes reasonably fast. Go may become an alternative, but it's a little too weird to go mainstream yet. C++ has turned out to be a mess. It adds hiding to C without adding memory safety, an unfortunate feature combination unique to C++.

Realistically, a CS degree today needs to cover machine learning, which is all about calculus and matrix math. There's less need for discrite math and bit-pushing.

I have classic CS training - all that stuff in vol. 1 of Knuth, automata theory, optimization of logic gates, formal methods, proof of correctness, etc. It's just not that useful any more. Mostly I write Python and Javascript.