On the other hand, an exception could be made on the grounds that it would make electric cars more common, which would be a net gain for the environment even with a polluting factory. This really doesn't sound like they're using this justification, but it's a possible one.

It's falling prices, but it's a measure of how fast they're falling. Not too long ago, $1/GB was the "barrier" everyone wanted to cross. Before that it was probably $5/GB or something. Next we'll be looking to break $0.25/GB, then probably price parity with hard drives.

It's like the 1GHz barrier on CPUs, back in the day. It wasn't so much a barrier as it was a milestone, a mark of how far we've progressed.

I skimmed TFA to find the actual math she earned it for. The summary they give is actually pretty interesting, even though they don't go into much detail on the math. Definitely doesn't seem like a bullshit hey-look-we're-giving-awards-to-minorities-too-now award.

This seems to be the actual paper, although to be honest it's so far above my knowledge that it could be about something completely different and I wouldn't be able to tell.

Yes, but some of them take a while to get the joke because their TLB had to be disabled.

(Certain releases of the "Barcelona" Opterons had a bug that could lock up the system. A workaround would prevent it, but had a stiff performance penalty. Later steppings had it fixed.)

Let's see...
Ariane 1 - second and fifth launches failed
Ariane 2 - only 6 launches, first failed
Ariane 3 - fifth launch failed
Ariane 4 - eighth launch failed
Ariane 5 - first launch failed, two partial failures in first 11
Atlas A - only 8 launches, 5 failed
Atlas B - only 10 launches, 3 failed
Atlas C - only 6 launches, 2 failed
Delta - first launch failed
Delta II - first eleven successful
Falcon 1 - only five launches, first three failed
Falcon 9 - first eleven launches successful, although a secondary payload on the fourth launch was aborted as a precaution
Long March 1 - only 2 launches, both successful
Long March 2 - first launch failed
Long March 3 - no complete failures in first 11, but 1 and 8 were partial failures
N-1 - only four launches, all failed horribly
Proton - third launch failed
Proton-K - second, third, fourth and sixth launches failed
Proton-M - eleventh launch failed
Saturn I - only ten launches, all successful
Saturn IB - only nine launches, all successful (unless you count Apollo 1 - it didn't launch but still killed three astronauts)
Saturn V - second launch (Apollo 6) failed, Apollo 13 doesn't count because it was a payload, not launcher, failure
Soyuz - third launch failed, with fatalities
Soyuz-U - seventh launch failed
Soyuz-FG - first eleven launches successful
Space Shuttle - first eleven successful (19th was first partial failure (ATO), 25th was first full failure)
Titan I - fifth, sixth, eighth, ninth and tenth launches failed
Titan II - ninth and eleventh launches failed
Titan III - first and sixth launches failed
Titan IV - seventh launch failed
Zenit-2 - first and second launches failed

Yep, getting zero failures in your first eleven launches is pretty damn rare.

In the letter, they keep going on about anomalies. They don't understand what those are.

Anomalies are not (necessarily) defects, or errors, or problems. Anomalies are deviations from the norm - something that isn't perfect.

I tried to find an example Space Shuttle mission that I could use to compare, but I can't even find a comprehensive list of "anomalies". I can find rollbacks, where the problem required bringing the vehicle back to the assembly building, but I can't find a list even of countdown stops.

Rockets are expensive. When you see a potential problem, you fix it even if there's a 90% chance of it being fine anyways. You don't take risks. For SpaceX, their caution has paid off in a near-100% success rate (one secondary payload was lost after an engine failed on CRS-1. NASA forbade the second burn to insert the secondary payload because the engine failure had reduced the odds of success to 95%).

Further, these are civilian launch vehicles, not missiles. A missile, you expect to be high-reliability, low-maintenance and weather-tolerant. You can't cancel a battle just because a hurricane is coming and you're not sure it can stand up to the wind. But these are civilian rockets - the increased payload and decreased cost you get from not having to battle-harden everything is worth the cost of having to delay the launch if something looks a bit iffy and they want to make sure it's not going to break and wreck your multi-million-dollar payload.

Oh, and then they somehow argue that having several billion dollars worth of flights sold is a bad thing. They frame it as "SpaceX is too slow to keep up with demand", when really it's "the demand is too high for SpaceX to keep up". They have missions sold out to 2019, and on many of them the payload isn't even ready yet. Replace SpaceX with even a perfect ideal, with an infinite supply of ready-to-launch rockets, and those seven Iridium-NEXT launches won't be happening until the actual payloads are done, the next five ISS resupply missions won't happen until the ISS needs the supplies, and the Falcon 9 Heavy test launch won't happen until that rocket is ready.

Let's think of it from another perspective - physical strength. Genetics does play a part in this, obviously, but training and exercise is easily the dominant factor.

Hair and eye color does not change significantly based on life experiences. Physical strength and intelligence does. For intelligence in particular, my experience has been that any genetic influence is almost unmeasurable compared to training and experience (this is in part because we have no good objective measure of intelligence).

I still think RT could have worked as a corporate tablet. Make it integrate 100% with AD and everything, give it built-in Office, and give bulk discounts when buying over a dozen of them. But no, Microsoft was blinded by the dreams of the consumer market, despite it being owned by Android/iPad, and so they missed the one niche they really could have nailed with it.

Here I am, in downtown Richmond (capital of Virginia). I *should* be getting some blazing-fast internet, right? Perfect conditions for it.

Nope. 3Mbps DSL. I can't switch ISPs because my apartment gave a monopoly to Telcom Communications (seriously, that's their actual name - they seem to be reselling CenturyLink). Sure, they don't call it that, but I checked every ISP and none of them will provide service to me except some DSL that's just as slow as what I've got.

And yet my parents, living twenty minutes away from anywhere in the empty part of Chesterfield, are getting 50Mbps FttH. I really want to see the economic explanation for that - it's too expensive to run fiber literally a block from Main Street, but a 20-mile run past several farms and lumber fields is somehow profitable.

Quite odd how, out of the first eighteen comments (not counting replies), five are about decommissioning costs, and five are about meltdowns? They seem to repeat the same talking points, almost as if on a script.

I'm not saying they're shills, but at the very least a lot of people seem to be getting their information from the same place, which leaves them missing several crucial facts:

1) Nuclear power works at scale. It's proven, and it scales perfectly. The biggest solar plants on the planet are 500MW (Topaz Solar Farm, PV) or 400MW (Ivanpah Solar Power Facility, thermal). A single nuclear reactor is well above that - scroll down this list and you'll see very few sub-500MW, and quite a few 1GW+ reactors. And remember, most plants have more than one reactor. 66 nuclear plants are enough to give us 20% of our energy. 947 wind plants are only enough to give us 3%, and 553 solar plants (PV and thermal) don't even break half a percent.

2) Nuclear power would be a hell of a lot safer if new designs were actually approved. The regulations are pretty much ridiculous - they don't approve new reactor types that are designed to solve all the problems we've found with the old designs, but they still allow old designs with known weaknesses to be extended long past their designed lifespan. Add to that the ridiculous costs of dealing with the bureaucracy and the weak requirements for cleanup/decommissioning, and it almost seems like the regulations are designed both to make nuclear power unprofitable, and to keep public opinion against it. Hmm...

3) Nobody is arguing for pure nuclear power, because that doesn't work for all the reasons people say it doesn't work. Nuclear (and geothermal, where possible) makes for an excellent base load. Nuclear meshes well with hydro - excess capacity can be used to run the dam in reverse, pumping water up to store that energy for later use. And if positioned right, it provides both cooling water for the reactor, and a single point to close off flow or install filters if something does go wrong. Wind, tidal and solar can supplement this as locations allow, with solar in particular taking the edge off the peak load.

4) Every power plant can go wrong. What happens when a hydro dam fails? Thousands of people die. What happens when a solar plant fails? We don't know yet, but it probably won't be that good considering how much damage they can do even when working properly. Same for wind, and tidal, and geothermal. They do some minor damage even when working perfectly - frying or chopping up migratory birds or fish, or altering the geology in the case of geothermal. Nuclear has the benefit, at least, of being perfectly clean when working perfectly. Yes, if things go wrong it can be absolutely horrible, but that's why regulations need to focus on redundant containment and fail-safe designs, not on constant inspections.

Alien Swarm: No sequels
Counter-Strike: one remake (Source), two sequels (Condition Zero, Global Offensive)
DOTA 2: No sequels, sort of a sequel itself though
Day of Defeat: One remake (Source)
Deathmatch Classic: No sequels
Half-Life: Three expansion packs (Opposing Force, Blue Shift, Decay), one remakes (Half-Life: Source), one deathmatch version (HL:DM), one remake of the deathmatch mode (HL:DM:S), one direct sequel (2), one sequel to the deathmatch mode (HL2:DM), three expansions to the sequel (Lost Coast, Episodes 1+2)
Left 4 Dead: One sequel (2)
Portal: One sequel (2)
Ricochet: No sequels
Team Fortress: One sequel (2)

This isn't counting the multiple non-Valve licensed variants of Counter-Strike, or the arcade version of Half-Life 2.

The image I had in my head, for some reason, was the driver leaving the car itself, which drove off without him. Apparently he wasn't stupid enough to do that, which is unfortunate because that would have been absolutely hilarious.

Not too much - it's one of those exponential curves that's shallow near the equator but steep near the poles.

Escape velocity is 11,186m/s. The ISS is at 7,650m/s. Keep those numbers in mind for a sense of scale..

At the equator, you get an extra 465m/s of velocity. At the poles, you get zero.

Boca Chica Village is at 25N. If I did my trig right, you'll get 420m/s of "free" velocity from a launch there.

For more comparison, Canaveral (28N) gets 410m/s, Wallops (38N) gets 365m/s, and Baikonur (46N) gets 320m/s of boost.

San Juan, Puerto Rico, is at 18N, which would get you 440m/s. A 20m/s difference, at the cost of shipping your rockets and payloads across the ocean, and building substantially more infrastructure. The economics does not support building a spaceport there.

So it's more like a Space Coast Guard than a Space Navy?

Almost - you also need to count memory channels, which on most desktops is two.

2 channels * 64 bits/channel * 2133MT/s / 8 bits/byte = 34128MB/s = 33.3GB/s

GPUs tend to use very wide, high-speed memory, because they need a lot more bandwidth than CPUs because graphics stuff doesn't cache as easily. Some data for comparison:
R7 240: 2 channels * 64 bits/channel * 4500MT/s = 70GB/s
R7 260X: 2 channels * 64 bits/channel * 6500MT/s = 102GB/s
R7 270X: 4 channels * 64 bits/channel * 5600MT/s = 175GB/s
R9 280X: 6 channels * 64 bits/channel * 6000MT/s = 281GB/s
R9 290X: 8 channels * 64 bits/channel * 5000MT/s = 312GB/s