But, how can this argument be made more compelling? I mean, for example, is there an example of a knockdown kit from abroad you might point to?
To my biggest surprise, when I left a minor outlet (epam) for Luxoft for a 3 month gig, I was very surprised to see the same TMobile, with the very same project, but similar specs. Lib A should do that and this, method C and D, must pass test, E and F.
Back in 2009, the few Google Russia people I knew told that task they get from HQ were almost like competitive coding tasks, pull record A from using hash table index B, do that under N cycles under worst case scenario C,D, and E. And all with near nil human contact. Nobody could've said it for sure if the task is even physically doable. And for that particular task, they laboured for half a year of trial and failure, after exhausting every trick from computer science textbooks.
In the end they did make an index traversal algo that did it. What was unsettling them was the seemed disregard for the amount of resources spent, and no concern for "if it is really doable?," as they were getting from supervisors the same answer all the time: "if you think it's impossible, not you, but somebody else eventually will do it, I am 100% sure"
Note: https://qz.com/india/1152683/indian-it-layoffs-in-2017-top-5...
The few people from SV crowd I can speak frankly with, said "saying that will certainly not make you more popular here."
Do people realise that these "mules" are the real heroes here who do all the hard work on algorithmics. In a decade or so, USA based companies will certainly begin to scratch their heads.