It looks to me like the solution to the problem is simple. Just have a "high tech visa" with no number limit, but one simple requirement: The company requesting the job needs to be offering something like 150% (or 200% or whatever makes sense) the average salary of the US workforce, or in their state.
This will ensure that the visa will only be offered for skills that are difficult to find in the USA (or the specific State). No problem of driving salaries down.
This is a great suggestion, especially if we accept the authors suggestion that hiring agents are only interested in exceptional foreign programmers, not the merely competent ones.
Also, what's a good ratio of exceptional programmers to competent programmers within an organization? Not all important code needs to be written by a clever architect.
This is essentially how the 457 visa works in Australia. It's uncapped but you have to pay a 'market salary rate', which is either the salary of an equivalent worker in the firm, or of an equivalent worker In the industry.
The problem with what you mention about Australia is regression to the mean. 'Market salary rate' means average salary, which is not what a great programmer is worth.
Actually, the state should also be irrelevant. Because if you can't find a programmer in your state then there's the rest of the country. The 150% idea is great -- it would create an incentive to only try to hire the exceptional, which is what PG wants right? I'm unconvinced; it feels like this is just a tech leader attempting to lower the cost of labor as opposed to increasing the quality -- otherwise he'd be promoting maximizing the quality of the 100k already coming in as opposed to increasing tr raw number of entrants. The only benefit to increasing supply is to lower wages. "Finding the exceptional" is a Trojan horse for lowering costs. As I've mentioned before, he's advocating a strip mining approach as opposed to a more surgical improvement in the quality of imports. Perhaps another means is to have a fixed number of visas as we do now, but auction them off to companies. The sweatshop companies wouldn't benefit because it would cost more than it saves and the other companies would then be far more selective, thus increasing the percentages on finding that 'exceptional' developer unicorn everyone seems to be chasing. I don't have the right answer, but I'm certain it isn't "let more people in and hope the 5% are among them."
an excellent suggestion, also possibly a bidding process... but I'm sure none of the US tech companies will champion this option because there is already room to do this unofficially within the current program... 50,000+ yearly visas gives a lot of overhead above just seeking "truly exceptional" talent.
Big tech companies can play the lottery, smaller ones can't.
And I don't see any need for a bidding process, which implies a fixed quota, a fixed closing date for the bidding, and a lot of uncertainty.
Just require the salary to be high enough to show you really need the skill, not to cut costs, and have a process with a guaranteed timing and outcome.
This will ensure that the visa will only be offered for skills that are difficult to find in the USA (or the specific State). No problem of driving salaries down.