I know about colocation, but I was under the impression that once you see an order in the book it is in the book. I.e. your order will be placed after it.
Leaving aside the environmental damage, which is obvious albeit somewhat sensationalized in this comment, how is said plastic actually harming human health today?
The article from Scientific America in the comment refers to the harm plastic can have in other animals in words such as "may", "might", "could". Has there been any study on how it is, now and actually, harming our health? Any measurements, case studies, or collective data from people suffering from any sort of disease related to plastic pollution?
Here in Brazil my experience has been the same: drivers use all apps.
I just don't see how Uber can maintain said network on the long run and still be profitable, not something I'd put my money in.
What Uber is doing would generally be described as predatory pricing on other sectors. On sufficiently unregulated markets such is Uber's it's extremely risky, no anti-trust legislation is needed to trump it, just a continuously low entry barrier, including lax regulation. For a more concrete, though only related example (rather than speculation as to Uber's future) see Dow's history.
Of course they're creating value, Uber and the like is of great value for its users.
The service is merely being subsidized by investors who believe in such practice.
Is it a bad investment? Maybe, their investors did not think so and they were free to compare it with other options you deem obviously better, considering you're even saying Uber and the like are stealing these other business would-be money...
IIRC this call has an overhead of a few ns. Isn't that close to or even higher than the time it takes to perform the action being measured on every loop? If so the author is just measuring clock_gettime. It can be confirmed on the author's system by simply calling clock_gettime without calling the measured action (movntdqa / cache flush) and comparing the results. An alternative approach is to call clock_gettime before and after the loop, not on every iteration, and then take an average.