The report quality is adequate to run of the mill federal cases of bomb threats and swatting investigations.
It is nice to see a nod given to SA in the report (though it does not go into much detail) - goons, true to their name, were the ones often pitting reddit/chans vs wu/sarkeesian against each other (by false flagging both camps) just for the sake of spectacular flamewars such internet cockfights tend to generate.
Short sightedness and excessive risk aversity are consequences, I'm interested in inherent causes of it. Calling it "greed" is a bit too nebulous - there's always profit motive, regardless of market strategy.
Gigabit to entire towns is a loss-maker not just in terms of decades.
The long haul costs are pretty trivial these days, the problem is still mostly last mile. Not in terms of technology, but legislation, leases, slush fund to get dig/pole permits and more, and that is assuming there are no state ordinances explicitly foiling metro net development like it is common in the US.
It indeed is similiar to real estate development in many ways. Needless to say UK has one of the most fucked up real estate markets, no surprise last mile competition is effectively nil.
Any infrastructure like that is basically loss-making for DECADES
Northern, central and part of eastern europe beg to differ, ROI for city is 3-6 months/cust, distant rural areas about 8-16 mo/cust. A lot of ISPs cover that by contract clauses, so they are guaranteed to get their break even money always.
City metro telcos + high speed wifi routinely cover 30% of population. The secret to it is commieblocks and just-in-time infra deployment - not sure how exactly that is viable in the UK, but in here, FTTX can be done for very cheap in densely populated cities, and those cover good chunk of population.
Remote areas get internet 2x more expensive, but still very cheap - mostly as a consequence of cut-throat urban competition, ISPs are forced to branch out into rural areas.
What happened to smart investors going long for decade spans?
Times too uncertain for long bets?
With google fiber it is even more bizarre. It is widely known that last mile telco returns are investments which routinely break even in a decade, there should be surprises about the initial capex. Not to mention that google completely ignored the synnergic effects.
In fact, a change to the government employment reporting came under Obama in 2010, as we're now using a different definition to calculate the "labor force", excluding people who have given up on finding work entirely, if I remember correctly. That changed the way we'd been measuring unemployment for the past thirty years.
This is incorrect anti-obama propaganda, and prime example of 'alternative facts'. Excluding 'people who have given up on finding work entirely' - that method is in place far, far longer. Obama didn't suddenly invent that in 2010. It is true that BLS changed the definition of "in labor force, seeking employment" in 2010 - but by extending the time limit to seek for job from 2 years to 5. Meaning if anything, the number is skewed towards more % unemployed, as now the time window of being considered "jobless" is bigger.
If you want to look for real fact fudging done by obama administration, look for downplaying of underemployment (ie shit dead end min wage jobs) instead.
As for raegan suddenly introducing that metric as implied in the TFA seems like equal bullshit, this time anti-reagan propaganda. The BLS metric was used at least since 1950.
Source: http://usatoday30.usatoday.com...
The crucial, extremely illegal link: https://bitbucket.org/prestoco...
And it compiles on modern linux, whoa. Can we have opensource, non-webkit, non-gecko browser now? Seriously, the slowness and resource of post-opera browsers is getting ridiculous - explorer is the fastest explorer at this point (not by js benchmarks, but overall user experience is far snappier).
Backward compatibility is what made microsoft originally great. I mean seriously MS, planned obsolescence? What the fuck happened to gratious phasing out?
You broke d3d9 in win10, horribly (much worse performance, different behavior leads to rendering artifacts). Reasoning that people would suddenly switch to UWP when everything runs like crap or something. Well thank you, if I wanted shoddy legacy API emulation, I can just run wine.
But the body of software using dx9 is still immense (to be xp compat, you need to use dx9). And this is just one of dozens of such examples. While w10 brings a lot of good innovation - it's generally leaner and faster than win7 - it also irrebarably broke key subsystems making a lot of software made before 2014 unusable on windows 10.
All modern VM solutions support snapshotting (let's disregard zones/jails/lxc/openvz containers which work a bit differently, but the fundamental problem is the same - host can do something without guest ever knowing).
Snapshots is exact replica, both disk, ram and system state of a live guest system, with that data it can be resumed - "forked" anew somewhere else, the guest never "reboots". When LEA writes a gag order for a VPS, they no longer ask for mere disk images, but for a live VM snapshot. Snapshots work atomically, without the guest system ever knowing. Normally it is used for live system migrations, load balancing and more.
Wrt encfs, same situation as dmcrypt - are you keeping it mounted, or not? If it is mounted, the raw encryption key ends up in the snapshot - OS has to be able to access the disk somehow.
As for ssh, it decrypts private keys with passphrase only momentarily during auth phase, but keeping keys on the server as such (even if passphrase protected) is really bad idea anyway (what if the box gets actively backdoored to snoop passphrases?). Just use agent forwarding where the server becomes literally a 'dumb pipe'.
Good luck finding one that will allow you to encrypt your whole partition.
Pretty much all providers with KVM technically support this, whether they intend it or not.
The problem is more fundamental. With VPS (as apposed to physical box where one can be certain that RAM snapshots are not easily made) - encrypting a partition won't help, just waste the already constrained cpu cycles. The provider still can trivially make a snapshot of your running VM, including live keys in memory, so what's the point? This is why businesses are not particularly fond of cloud things, and tend to replicate that stuff in-house. There's a field of ZK utility computing, but it is still quite inefficient.
Look at the codec diagram - if you ignore the entropy coder, it largely resembles input filters of voicerecog systems - before feeding the NN input terminals, signal is decimated to extremely low bandwidth vectors with only the psychoacoustic essentials of human voice - quantized to very few dominating tones and their attack/release values. The NN model does the final step of "compressing" the result only by factor of around 100 into text. It is popularly conjenctured that compression is, in fact, a ML problem.
Same is done with computer vision, before matching for features, the frequency space is filtered into a narrow band where the interesting stuff can be still observed.
It's interesting how the polarizing emergence of PC and neoreactionary extremes roughly correlates with the time fb started to bias post order according to personal bubbles (facebook started forcing "algorithmic feed" in 2010). Could be mere coincidence, then again, it may be not. In which case, censoring hoaxes would be merely cosmetic band aid, doing nothing to fix the spiral of ignorance facebook could be causing in the first place.
Let's skip the boring technical details, and ask some juicier insider info.
Apple has done great things for LLVM/Clang. I believe you were the key conductor in this. Does it mean you leaving we can expect less in this regard in the future, or are you just passing the torch in there, keeping it capable hands (aka there's not much left for you to do)?
There was never a competition, even though it compiles, swift does not abstract android APIs to any reasonable measure to this day so it is extremely inconvenient in practice.
Even xamarin is far from a silver bullet - it is extremely mobile specific, yet has many even more popular competitors (various js-based stuff, corona is very popular too).
No doubt about RNC it is severely tainted since the Bush era, especially hawks. However that's the thing about Trump - he did everything to *not* be seen in the same bed with RNC - meaning disclosing dirt on RNC would be just purposelessly burning political capital (plus, dems used to pride themselves to not "play dirty" like that), as Trump would come clean from such attacks.
The asymmetry lies elsewhere - not in disclosure, but existence of the compro as such - Trump was largely unrelated to big politics cliques aside from trivial stuff such as bribes to get building permits and such, he had no *chance* to keep dangerous political skeletons in the closet. The only dirt they could find on him is more or less the usual you could find on any celebrity of his caliber - which pales in comparison compared to people who had access to real power for years.
Indeed it's overdramatizing stating that DNC had their own people there directly - however, that D leaning media (ie CNN, NYT, The Hill..) gave Clinton preferential treatment at the expense of their journalist integrity is no secret, it was widely covered during the election even by neutral media such as politico. Interestingly, right after that those same media started pushing the narrative these emails are "planted documents" by Russia.
Yet you can't be bothered to explain this alleged complexity. My view is that it really is that simple.
Realistic conflict theory is much more than just "duh, ig/og". The complexity is of why competing cliques are formed in the first place - the explanations in terms of racial/gender (both left and right) theories appear indeed simple, but biased to the point those also completely contradict empirical research in social sciences.
But when you start picking and choosing which ethnic groups to protect while other ethnic groups aren't even recognized, it's quite clear that's the usual in-group/out-group dynamics going on.
I agree it's a problem. Right wingers are concerned about reverse racism while social left is concerned that protecting classes does not protect certain groups sufficiently (thus you get AA, which antagonizes right, in endless cycle).
You're correct that protected class is simply protection of attribute as such, regardless of value of that attribute. Can't discriminate based on sex or ethnicum, regardless of what actual gender or race one is.
Here you go with assumptions again. It's more complex than in-group/out-group. We're not talking targeted aggresion here, but mere caricature - a re-enactment of stereotype for comedic effect. For example some people took issue with this skit, despite the trope (90s TV advertising) being spot on - or a more recent one.
As long it's just bunch of people yapping I don't really care, but the moment there's a chilling effects because of this, it means a lot of people completely missed the point of having protected classes.
tl;dr car analogy: campaigning to ban driving because people die on roads is just plain overreacting. chill out, people.
I assume the point is that caricaturing men is funny, while caricaturing women is berating. I never really quite understood why people obsess about it so much anyway, there are thousands such double standards in life. Rationalizations of unjust world such as "feminazis", "internalized patriarchy" and "microaggresions" sounds like hearing children who never learned to cope with being adults.
1) No botnet actually hijacked 900k CPEs of DT, at the moment there are rougly between 10k-40k zyxel ones across the world. The outages were caused by the increased 7547 scan traffic crashing routers of other vendors.
2) Zyxel SOAP RCE probes died down rapidly past 2 weeks. There is still some traffic (wget vizxv.pw/a if you're curious, note that you need actual wget user-agent), but the botnet is relatively small at this point.
3) As for general IoT botnets using telnet, running a simple cowrie honeypot will tell you that C&C method of current largest botnet is not Tor based, but bittorrent DHT based. The codebase appears to be unrelated to mirai, too.
All of the above can be fact checked using pretty simple tools - for TR-069 exploit simply listen with netcat, for telnet/ssh bruteforce use cowrie. Botnet size can be gauged accurately by sampling scan probes (mirai codebase sends 160 probes/s).
Next should be currency that any of us are allowed to create
Already the case with most of keynesian currencies, aka elastic supply. While not really "anyone" can "print" the IOUs, when borrowing money, reserve banking "prints" those out of thin air.
The only way to make it even more fair would be simply scrap the IOUs and print naked (which is what "negative interest rate"s pretty much boil down to) - but as a basic income.
This basic income/QE infinity hybrid will coincide with immense deflationary pressure (already somewhat present) because of capital concentration and people being pissed that majority of cheap credit ("printing") currently ends up fueling illiquid ponzis such as housing bubbles and corporate buybacks, instead of consumer economy.
Uber dictates their working patterns once they have logged on, has raised its commission while cutting the rates they can charge, and imposes lockouts from its system if drivers turn down too many jobs.
Translation:
Market exchange dictates the trading hours, imposes ridiculous trading commision fees (15-30%), puts a cap on the ask offers and kicks you out if you don't execute enough trades.
Now, why on earth would sellers stay on a market this shitty? Bandwagon effect. Other competing exchanges don't have the liquidity. Why people use Microsoft products? Bandwagon effect. Once you get something shitty going, it can keep going on its momentum alone.
That the exchange can dictate price levels really is a problem because it creates race-to-the-bottom pressures - negative feedback loop - drivers can't go to competing markets which treat em better, because their cheap labor keeps those alternative companies out of the business (and even if those adopt similiarly shitty business practices, they end up being no better than uber). Thus the accusations of entrapment.
If Uber wants to be merely a clearing house for car hailing settlements that's fine, but people should call it out on their attempts to corner the market in order to keep their first mover monopoly.
Na. Start with TAOCP, then read the harder books (SICP, CLRS).
I can't read math for shit either, and it's not really that necessary for the verbose style of knuth. So one can do just fine with just pseudocode.
Why SICP later? It is a bit harder to read than TAOCP, as it is not really a bag of ready to use tricks anymore - it challenges the reader to think about (functional) programming at a more fundamental level (and you don't need to know much about "math" either, it's an intro course book).
Err, make that 70grams. This is because probability of atom decaying in half-life is, well, half, 50%.
Also that ballpark can be way off by magnitude or three, either due to my error or inefficiences in the electron capture. Even if its 4 magnitudes off, it's still very competive with contemporary RTGs which are limited in efficiency by costly radiators.
You don't want to care about energy density (aka capacity), as that number is insane for anything nuclear.
You do want to know the internal resistance, rate of "discharge", basically watts it can produce for given weight. You can burn uranium or gasoline in an instant, but decay mode sources (RTG and this) are limited to rate of decay.
Back of envelope:
(all exponents are to power of 10, not 2).
One C-14 atom decays in 5730 years, shoots off 156476 electron volt we ideally capture, and one anti-neutrino we shield off
One electron volt is 1.6e-19 J, ie that beta decay is roughly 2.5e-14 J.
You need 10^14 atoms of c-14 to get 2.5 joule every 5730 years.
Now for watts (aka joule per second):
5730 * 360 * 86400 * 10^14 to get 2.5 watts output
5730 * 360 * 86400 * 10^14 / 2.5 =
7.1e24 C-14 atoms per watt.
Now C-12 atom supposedly weights 2e-23 grams and C-14 should be in same ballpark, meaning
35 grams of this stuff, in ideal case gives off 1watt,
Which is pretty impressive, for a battery which lasts basically forever.
Apple out of world markets due to not being able to be price competitive.
Nah, Apple is one of the few exceptions. It would eat into their profits a small bit, but their business definitely does not depend on razor thin market margins on hardware, its the opposite. Same could be said about a lot of SV companies.
The problem isnt cost of manufacture, but competing for engineering talent. If Trump stomps on their H1B lobby, SV competiteveness compared to rest of the world will be severely impacted.
Slashdot Mirror
The report quality is adequate to run of the mill federal cases of bomb threats and swatting investigations.
It is nice to see a nod given to SA in the report (though it does not go into much detail) - goons, true to their name, were the ones often pitting reddit/chans vs wu/sarkeesian against each other (by false flagging both camps) just for the sake of spectacular flamewars such internet cockfights tend to generate.
Short sightedness and excessive risk aversity are consequences, I'm interested in inherent causes of it. Calling it "greed" is a bit too nebulous - there's always profit motive, regardless of market strategy.
Running a company isn't a netflix binge.
The long haul costs are pretty trivial these days, the problem is still mostly last mile. Not in terms of technology, but legislation, leases, slush fund to get dig/pole permits and more, and that is assuming there are no state ordinances explicitly foiling metro net development like it is common in the US.
It indeed is similiar to real estate development in many ways. Needless to say UK has one of the most fucked up real estate markets, no surprise last mile competition is effectively nil.
Northern, central and part of eastern europe beg to differ, ROI for city is 3-6 months/cust, distant rural areas about 8-16 mo/cust. A lot of ISPs cover that by contract clauses, so they are guaranteed to get their break even money always.
City metro telcos + high speed wifi routinely cover 30% of population. The secret to it is commieblocks and just-in-time infra deployment - not sure how exactly that is viable in the UK, but in here, FTTX can be done for very cheap in densely populated cities, and those cover good chunk of population.
Remote areas get internet 2x more expensive, but still very cheap - mostly as a consequence of cut-throat urban competition, ISPs are forced to branch out into rural areas.
What happened to smart investors going long for decade spans?
Times too uncertain for long bets?
With google fiber it is even more bizarre. It is widely known that last mile telco returns are investments which routinely break even in a decade, there should be surprises about the initial capex. Not to mention that google completely ignored the synnergic effects.
This is incorrect anti-obama propaganda, and prime example of 'alternative facts'. Excluding 'people who have given up on finding work entirely' - that method is in place far, far longer. Obama didn't suddenly invent that in 2010. It is true that BLS changed the definition of "in labor force, seeking employment" in 2010 - but by extending the time limit to seek for job from 2 years to 5. Meaning if anything, the number is skewed towards more % unemployed, as now the time window of being considered "jobless" is bigger.
If you want to look for real fact fudging done by obama administration, look for downplaying of underemployment (ie shit dead end min wage jobs) instead.
As for raegan suddenly introducing that metric as implied in the TFA seems like equal bullshit, this time anti-reagan propaganda. The BLS metric was used at least since 1950.
Source: http://usatoday30.usatoday.com...
The crucial, extremely illegal link: https://bitbucket.org/prestoco... And it compiles on modern linux, whoa. Can we have opensource, non-webkit, non-gecko browser now? Seriously, the slowness and resource of post-opera browsers is getting ridiculous - explorer is the fastest explorer at this point (not by js benchmarks, but overall user experience is far snappier).
Backward compatibility is what made microsoft originally great. I mean seriously MS, planned obsolescence? What the fuck happened to gratious phasing out?
You broke d3d9 in win10, horribly (much worse performance, different behavior leads to rendering artifacts). Reasoning that people would suddenly switch to UWP when everything runs like crap or something. Well thank you, if I wanted shoddy legacy API emulation, I can just run wine.
But the body of software using dx9 is still immense (to be xp compat, you need to use dx9). And this is just one of dozens of such examples. While w10 brings a lot of good innovation - it's generally leaner and faster than win7 - it also irrebarably broke key subsystems making a lot of software made before 2014 unusable on windows 10.
All modern VM solutions support snapshotting (let's disregard zones/jails/lxc/openvz containers which work a bit differently, but the fundamental problem is the same - host can do something without guest ever knowing).
Snapshots is exact replica, both disk, ram and system state of a live guest system, with that data it can be resumed - "forked" anew somewhere else, the guest never "reboots". When LEA writes a gag order for a VPS, they no longer ask for mere disk images, but for a live VM snapshot. Snapshots work atomically, without the guest system ever knowing. Normally it is used for live system migrations, load balancing and more.
Wrt encfs, same situation as dmcrypt - are you keeping it mounted, or not? If it is mounted, the raw encryption key ends up in the snapshot - OS has to be able to access the disk somehow.
As for ssh, it decrypts private keys with passphrase only momentarily during auth phase, but keeping keys on the server as such (even if passphrase protected) is really bad idea anyway (what if the box gets actively backdoored to snoop passphrases?). Just use agent forwarding where the server becomes literally a 'dumb pipe'.
Pretty much all providers with KVM technically support this, whether they intend it or not.
The problem is more fundamental. With VPS (as apposed to physical box where one can be certain that RAM snapshots are not easily made) - encrypting a partition won't help, just waste the already constrained cpu cycles. The provider still can trivially make a snapshot of your running VM, including live keys in memory, so what's the point? This is why businesses are not particularly fond of cloud things, and tend to replicate that stuff in-house. There's a field of ZK utility computing, but it is still quite inefficient.
Look at the codec diagram - if you ignore the entropy coder, it largely resembles input filters of voicerecog systems - before feeding the NN input terminals, signal is decimated to extremely low bandwidth vectors with only the psychoacoustic essentials of human voice - quantized to very few dominating tones and their attack/release values. The NN model does the final step of "compressing" the result only by factor of around 100 into text. It is popularly conjenctured that compression is, in fact, a ML problem.
Same is done with computer vision, before matching for features, the frequency space is filtered into a narrow band where the interesting stuff can be still observed.
It's interesting how the polarizing emergence of PC and neoreactionary extremes roughly correlates with the time fb started to bias post order according to personal bubbles (facebook started forcing "algorithmic feed" in 2010). Could be mere coincidence, then again, it may be not. In which case, censoring hoaxes would be merely cosmetic band aid, doing nothing to fix the spiral of ignorance facebook could be causing in the first place.
Let's skip the boring technical details, and ask some juicier insider info.
Apple has done great things for LLVM/Clang. I believe you were the key conductor in this. Does it mean you leaving we can expect less in this regard in the future, or are you just passing the torch in there, keeping it capable hands (aka there's not much left for you to do)?
Or some other change of circumstances?
There was never a competition, even though it compiles, swift does not abstract android APIs to any reasonable measure to this day so it is extremely inconvenient in practice.
Even xamarin is far from a silver bullet - it is extremely mobile specific, yet has many even more popular competitors (various js-based stuff, corona is very popular too).
No doubt about RNC it is severely tainted since the Bush era, especially hawks. However that's the thing about Trump - he did everything to *not* be seen in the same bed with RNC - meaning disclosing dirt on RNC would be just purposelessly burning political capital (plus, dems used to pride themselves to not "play dirty" like that), as Trump would come clean from such attacks.
The asymmetry lies elsewhere - not in disclosure, but existence of the compro as such - Trump was largely unrelated to big politics cliques aside from trivial stuff such as bribes to get building permits and such, he had no *chance* to keep dangerous political skeletons in the closet. The only dirt they could find on him is more or less the usual you could find on any celebrity of his caliber - which pales in comparison compared to people who had access to real power for years.
Indeed it's overdramatizing stating that DNC had their own people there directly - however, that D leaning media (ie CNN, NYT, The Hill..) gave Clinton preferential treatment at the expense of their journalist integrity is no secret, it was widely covered during the election even by neutral media such as politico. Interestingly, right after that those same media started pushing the narrative these emails are "planted documents" by Russia.
http://www.politico.com/story/...
Realistic conflict theory is much more than just "duh, ig/og". The complexity is of why competing cliques are formed in the first place - the explanations in terms of racial/gender (both left and right) theories appear indeed simple, but biased to the point those also completely contradict empirical research in social sciences.
I agree it's a problem. Right wingers are concerned about reverse racism while social left is concerned that protecting classes does not protect certain groups sufficiently (thus you get AA, which antagonizes right, in endless cycle).
You're correct that protected class is simply protection of attribute as such, regardless of value of that attribute. Can't discriminate based on sex or ethnicum, regardless of what actual gender or race one is.
Here you go with assumptions again. It's more complex than in-group/out-group. We're not talking targeted aggresion here, but mere caricature - a re-enactment of stereotype for comedic effect. For example some people took issue with this skit, despite the trope (90s TV advertising) being spot on - or a more recent one.
As long it's just bunch of people yapping I don't really care, but the moment there's a chilling effects because of this, it means a lot of people completely missed the point of having protected classes.
tl;dr car analogy: campaigning to ban driving because people die on roads is just plain overreacting. chill out, people.
I assume the point is that caricaturing men is funny, while caricaturing women is berating. I never really quite understood why people obsess about it so much anyway, there are thousands such double standards in life. Rationalizations of unjust world such as "feminazis", "internalized patriarchy" and "microaggresions" sounds like hearing children who never learned to cope with being adults.
1) No botnet actually hijacked 900k CPEs of DT, at the moment there are rougly between 10k-40k zyxel ones across the world. The outages were caused by the increased 7547 scan traffic crashing routers of other vendors.
2) Zyxel SOAP RCE probes died down rapidly past 2 weeks. There is still some traffic (wget vizxv.pw/a if you're curious, note that you need actual wget user-agent), but the botnet is relatively small at this point.
3) As for general IoT botnets using telnet, running a simple cowrie honeypot will tell you that C&C method of current largest botnet is not Tor based, but bittorrent DHT based. The codebase appears to be unrelated to mirai, too.
All of the above can be fact checked using pretty simple tools - for TR-069 exploit simply listen with netcat, for telnet/ssh bruteforce use cowrie. Botnet size can be gauged accurately by sampling scan probes (mirai codebase sends 160 probes/s).
Already the case with most of keynesian currencies, aka elastic supply. While not really "anyone" can "print" the IOUs, when borrowing money, reserve banking "prints" those out of thin air.
The only way to make it even more fair would be simply scrap the IOUs and print naked (which is what "negative interest rate"s pretty much boil down to) - but as a basic income.
This basic income/QE infinity hybrid will coincide with immense deflationary pressure (already somewhat present) because of capital concentration and people being pissed that majority of cheap credit ("printing") currently ends up fueling illiquid ponzis such as housing bubbles and corporate buybacks, instead of consumer economy.
Translation:
Now, why on earth would sellers stay on a market this shitty? Bandwagon effect. Other competing exchanges don't have the liquidity. Why people use Microsoft products? Bandwagon effect. Once you get something shitty going, it can keep going on its momentum alone.
That the exchange can dictate price levels really is a problem because it creates race-to-the-bottom pressures - negative feedback loop - drivers can't go to competing markets which treat em better, because their cheap labor keeps those alternative companies out of the business (and even if those adopt similiarly shitty business practices, they end up being no better than uber). Thus the accusations of entrapment.
If Uber wants to be merely a clearing house for car hailing settlements that's fine, but people should call it out on their attempts to corner the market in order to keep their first mover monopoly.
Na. Start with TAOCP, then read the harder books (SICP, CLRS).
I can't read math for shit either, and it's not really that necessary for the verbose style of knuth. So one can do just fine with just pseudocode.
Why SICP later? It is a bit harder to read than TAOCP, as it is not really a bag of ready to use tricks anymore - it challenges the reader to think about (functional) programming at a more fundamental level (and you don't need to know much about "math" either, it's an intro course book).
Err, make that 70grams. This is because probability of atom decaying in half-life is, well, half, 50%. Also that ballpark can be way off by magnitude or three, either due to my error or inefficiences in the electron capture. Even if its 4 magnitudes off, it's still very competive with contemporary RTGs which are limited in efficiency by costly radiators.
You don't want to care about energy density (aka capacity), as that number is insane for anything nuclear. You do want to know the internal resistance, rate of "discharge", basically watts it can produce for given weight. You can burn uranium or gasoline in an instant, but decay mode sources (RTG and this) are limited to rate of decay.
Back of envelope: (all exponents are to power of 10, not 2).
One C-14 atom decays in 5730 years, shoots off 156476 electron volt we ideally capture, and one anti-neutrino we shield off
One electron volt is 1.6e-19 J, ie that beta decay is roughly 2.5e-14 J.
You need 10^14 atoms of c-14 to get 2.5 joule every 5730 years.
Now for watts (aka joule per second): 5730 * 360 * 86400 * 10^14 to get 2.5 watts output
5730 * 360 * 86400 * 10^14 / 2.5 =
7.1e24 C-14 atoms per watt.
Now C-12 atom supposedly weights 2e-23 grams and C-14 should be in same ballpark, meaning
35 grams of this stuff, in ideal case gives off 1watt,
Which is pretty impressive, for a battery which lasts basically forever.
Nah, Apple is one of the few exceptions. It would eat into their profits a small bit, but their business definitely does not depend on razor thin market margins on hardware, its the opposite. Same could be said about a lot of SV companies.
The problem isnt cost of manufacture, but competing for engineering talent. If Trump stomps on their H1B lobby, SV competiteveness compared to rest of the world will be severely impacted.