Paul Graham wrote Let the Other 95% of Great Programmers In. But I find it hard to believe that he got both programmer quality and supply & demand so wrong. Actually, that’s an understatement. I am angry that Graham argues, as a capitalist, that the US government should grant him exemption from supply and demand (of labor).

The body of his short essay includes the term ‘great programmer’ 8 times; a footnote has 5 more occurrences. In that footnote Paul implies that a ‘great programmer’ might be worth 100 to 1,000 times as much as a non-great programmer. That’s absurd, but let’s consider famous programmers for a moment (alas, Lady Ada Lovelace is not useful for salary comparisons). Mark Zuckerberg isn’t as valuable as 100 peers as a programmer. Having the idea for Facebook was valuable, but that wasn’t programming. That was a business/service insight (which several other people, famously, had around the same time). Marc Andreessen is not worth 100 of his peers as a programmer — and he’s an investor now. John Carmack is an obvious genius who keeps pushing the edge — but nobody ever hired him instead of 100 programmers.

Perhaps Paul is really thinking of product designers or systems architects, who do stuff normal programmers don’t do? But that’s not programming, and that’s not the programmer job that H-1B visas are used to fill. H-1B visas fill many jobs, and I doubt you could get one to be a startup entrepreneur. The US granted 85,000 H-1B visas in 2014. They weren’t for mad scientists, genius founders, or driven entrepreneurs. Most of those 85,000 people work alongside US Citizens, doing similar work for similar salaries. Those visas do not go to 85,000 prodigy rockstars in various fields.

I have screened and interviewed hundreds of systems administrators, both US Citizens and on visas. I work with sysadmins, programmers, and data analysts. We hire the best people we can get, but the problem is finding qualified people… which Graham apparently doesn’t care about. We don’t look overseas to find geniuses just waiting for visas and airfare. Aside from generally minor language issues, US Citizen vs. visa holder just isn’t relevant. Foreigners generally speak more languages than locals, who have better knowledge of local restaurants and movie theaters. Who cares?

The immigrants I interview and work with are not 100–1,000 times as effective as US Citizens, which is the implication I get from Paul’s article: that the US has plenty of non-great programmers, but we need to recruit outside our borders to find enough great programmers. Immigrants are not hired with an understanding or expectation that they will be twice as effective as US candidates. We hire immigrants (and employers deal with the costs and paperwork) because we need people to do lots of (often basic) jobs, and there are simply not enough qualified candidates — whether programmers, system administrators, or other tech types.

As a sanity check, do programmer salaries reflect Paul’s 100–1,000 multiple? A new programmer just out of school, without any real-world experience, might make $50k/year. If a great programmer can really be 100 times more effective than that baseline, they should be worth $5m/year (or more — that is relative to a starting salary). There might be a few Wall St. programmers who made $5m with bonus and options in a good year, but certainly not 85,000 of them each year. If visas were really for Paul’s “great programmers”, they should be getting $500k/year or more, and that is manifestly not the case — neither for US Citizens nor for visa holders.

I wonder why Paul and I see this so differently. One obvious difference between us is that Paul is an employer and I’m an employee.

The US has about 5% of the world population, but for argument’s sake let’s estimate we have 20% of the world’s high-tech jobs because we are so involved in pioneering technological development (including business models as much as basic R&D). I agree with Paul that this is a wonderful and valuable thing, which we should work hard to preserve. And of course it is ridiculous for Paul to claim that programmers are uniformly distributed among nations — to become a skilled programmers you need to be in a country with computers, electricity, time to focus on programming (rather than subsistence farming), and hopefully an educational system and job market where you can hone your craft.

In the US economy, leaving out the rest of the world for a moment, we are clearly not training enough high-tech workers to feed our labor demand. We have too many people coming out of school who cannot find jobs, and not enough tech types for Y Combinator (or my various employers) to hire. At the national level, supply and demand shows us the obvious fix: pay more. Pay more to hire programmers and tech types away from your competitors. In the short term, hiring programmers away from competitors would be painful for the competitors. But I keep hearing that disruption is great and unavoidable. Ride the wave or get crushed, right‽ I would enjoy that disruption.

So what’s a cheaper way for companies to hire tech types than letting demand set the market rate? Look harder. Expand your hiring scope outside the US. You can do that by sending jobs offshore — which is used mostly to hire cheaper labor, not to find great employees. Or by bringing guest workers into the US. But don’t pretend visas are the only way to hire the people you need. Or that H-1B visas are primarily for great programmers who are worth 100 times as much as regular programmers.

You don’t need visas to hire (or create) skilled employees. What would happen if a few large companies (let’s start with Apple, Google, Intel, and Adobe from the lawsuit; I’d add Y Combinator, Facebook, Twitter, Microsoft, IBM, and Oracle), all announced they would raise salaries in all H-1B categories by 5%/year for 10 years, while the US reduced H-1B visas by 10% per year for 9 years, ending at 8,500 (10% of current volume)?

1) All the tech companies would raise salaries and start aggressively ‘poaching’ from each other to fill demand for existing jobs.

2) Those companies would look long and hard at their hiring requirements and crank up their training to fill holes.

3) Anyone with a passing interest in technology would reconsider their employment prospects, and a lot more people would decide to focus on technology.

4) Across the US, parents would tell their kids to “learn computers”.

Yes, it would be expensive. H-1B visas are expensive. Recruiting is expensive. Employees are expensive!

As a bonus, though, higher salaries and less short-term visas would greatly alleviate the “reverse brain drain” of guest workers learning high-tech skills in the US (both in school and at work) and then taking their skills home to start companies which compete with their old US employers.

But be honest. H-1B visa demand is not high because companies are striving for excellence. The visas are being used to preserve the existing labor market (salary levels) rather than paying higher salaries as dictated by supply and demand. Paul’s suggestion would help US companies find employees, but drain brainpower from the rest of the world, and leave the US workforce uncompetitive for US jobs. It is not a good answer.

Thanks to @ftrain, @ksbrandi, @ceolaf, @adamengst, @rmogull, and Julia for advice on this piece.


I have seen many responses, both directly on this essay (numbered speech bubbles on the right), on Twitter, and elsewhere on Medium. I collected them on Storify: <>.