LLVM is probably too heavy-weight to run dynamically. It generates high quality code, but it isn't designed to compile short traces of instructions like you would need to do in a dynamic binary translator.

This might not be a bad idea if it could be done once when the binary is installed. However, this is just impossible to do in general for x86 because it isn't always possible to tell the difference between data and code before the program tries to execute it. This is especially true for applications that do any kind of runtime code generation themselves because they would be generating x86 (not ARM) instructions on the fly.

I'm somewhat sad that no one ships LLVM byte code binaries that are lowered to the native ISA at install time. It was one of the original design goals of LLVM, and it would have eliminated all of these issues.

This is a good post (the point about hiding memory latency in particular), but you should still wait to judge the new chip until benchmarks are posted.

If you have ever worked on a design team for a high performance modern CPU, you should know that high level classifications like OOO vs In-Order never tell the whole story, and most real designs are hybrids of multiple high level approaches.

This is a very commonly held belief, but it is not well supported by evidence. I would challenge you to question this belief and to read up on this topic.

"The causal effect of education on earning." by David Card and "Instructional Interventions Affecting Critical Thinking Skills and Dispositions." by Abrami et. al may be a good starting point.

As you point out, evaluating how much was learned by a student from these courses is a big problem. Existing standardized comprehensive examinations like the AP or SAT programs seem like at least a starting point for high school students to prove their abilities, but there isn't really anything equivalent for undergraduate students.

Do you actually believe this? Doesn't this attitude presuppose that a college education provides exactly zero value to students that would have otherwise gone to trade schools or just been happy with a high school diploma.

Perhaps you meant to say that there is some value provided, but it is diminished to the point that it isn't worth the investment in time or money for many students. I do agree with this, and my suggestion was that an alternative perspective on solving this problem could be to try to college education such that the value is increased, and the costs (time and money) are reduced. It's fair to be skeptical of whether or not this could be accomplished, but generally I think that it is better to attempt to try and fail to make progress on hard problems than to sit back and be sure that you will make no progress at all.

I do agree that many of these requirements are ridiculous and artificial.

However, I wonder if "educational inflation" is really detrimental in the long term. The main problem seems to be the increased cost of education (caused by higher demand) rather than the side effect of creating a more educated population. I wonder if these can actually be treated as separate issues, and effort can be focused on the former.

I'm genuinely jealous of younger students who have access to these resources. Of course not everyone is going to take advantage of them, but the options for independent learning were severely limited before these movements.

What do people think about the potential for these movements to address the growing costs of quality higher education? Of course Coursera and The Khan Academy are not good enough to be replacements for existing systems (yet), but what is missing and how can we close the gap?

You're entitled to your opinion, but I hope that I can convince you that the ideal number of merit based work visas should be greater than 0. If you really think that 650,000 is too many, then you should try to justify it by modeling the economic impact, and showing that a number significantly lower than 650,000 would have a significantly more positive impact on average salaries in the short term and the long term. I gave a crude estimate above that the number should be in the ballpark of 120,000 to 420,000. With roughly half a million jobs at stake, I assume that someone with expertise in economics and statistics has done this analysis and that the number of 650,000 was not accepted blindly without strong evidence.

My argument is that there should be an avenue for exceptional people to immigrate that is based on merit, not who you marry.

The problem is that it is hard to predict who is going to accomplish something noteworthy, so you have to cast a relatively wide net. You filter people out by their potential to do something great (merit), and the net result is still positive even if only a small fraction is successful.

How is the average wage relevant? Please explain.

The benefit derived from importing top talent should be independent of the current state of the economy. If the point of the program is to encourage new technology to be developed in the USA rather than somewhere else, then how is it a good idea to discourage domestic technology development during an economic downturn?

The reason why you shouldn't make decisions based on anecdotal evidence is that it is often biased. My anecdotal evidence says exactly the opposite, and I assume that it is biased.

I won't argue that there should be a limit, but I will argue that the limit should be high enough to absorb graduates from top science and engineering programs worldwide. Last year there were about 50,000 engineering masters degrees awarded and about 10,000 Doctorate degrees in the USA, the majority of which were international students (and even though not all of them are international students, I'll stick with these numbers as an estimate to account for graduates of top international schools that we could potential poach talent from). The transitional period between getting a degree and getting a green card ranges from 3-7 years, depending on the country. So collectively the immigration system should absorb something like 180,000-420,000 workers with temporary visas at any given time.

That's 0.2% of the population or 0.4% of the labor pool. To put it another way, it's 0.02% of the global workforce. Does that really seem so far off? That 1 out of 5000 workers in the world might have technical expertise that is worth importing into the US labor market?

Many companies will immediately start a green card application when you join, then you can extend your H1 until is processed.

Maybe they lied to me in grad school, but here's what I could find with a quick google search.

http://www.timeshighereducation.co.uk/world-university-rankings/2013-14/world-ranking

Of course they can, and for most jobs you are absolutely right. There won't be much difference if a job is given to an American vs an immigrant.

However, this isn't true at the very high end, particularly for individuals who create disruptive changes in technology. I will argue that we absolutely want those people immigrating to the USA.

To give an extreme example, imagine if the transistor was invented in England rather than Bell Labs (William Schockley was of English nationality and came to the USA on a student visa). What would the effect on the American economy be? I believe that it would eventually erode our leadership position in the technology industry and in education, which would negatively impact our economy.

Technology history is littered with examples like this, big ones that you have probably heard about (like the Viterbi Algorithm), as well as more modest advances that collectively advance the American economy. Take a look through the author list on recently issued patents or top-tier fundamental research publications in technology related fields, and you'll realize that most of these people were not born Americans.

The point of H1Bs is to cast a wide enough net that these people choose to work in the USA. Other than H1B, what avenues do these people have to obtain employment and residency in America?