While the Trial framework is capable of doing A/B testing (and historically started out for that purpose), it is fundamentally an asset delivery mechanism, and is mostly used for that (i.e. I suspect that most of the domains using Trial currently have no "B" to pair with their "A").
And some of the differentiation that IS being done is not A/B testing in the usual sense of the word. i.e. if one particular hardware model needs a different version of an asset, those models can be assigned to a separate set, but that does not really make it an A/B test.
Well, he would know about engineers being "retarded". I found him extremely difficult to work with, unwilling to accept that any behavior of his code was a bug, and an all around obnoxious guy.
Lattner was indeed THE lead programmer of Swift, but Gross was A Director working on Machine Learning not THE guy in charge of company wide machine learning efforts ("Director" at Apple basically means second level manager).
"In another sign that the company has prioritized the iPhone, Apple re-organized its software engineering department so there's no longer a dedicated Mac operating system team."
There are a number of problems with this line of reasoning:
1. That reorganization started years ago, in the wake of Scott Forstall's departure in late 2012
2. The previous organization was, in my opinion, much more detrimental to macOS, with the iOS organization operating as a roach motel, where code and engineers move into, but never back out again (My perception was that this was Scott's doing, but YMMV).
3. There is no longer a dedicated iOS operating system team either, so I don't see how that reorganization can be seen as deemphasizing macOS.
4. With iOS accounting for the vast majority of revenue, an arrangement to share infrastructure tends to be advantageous to macOS, if anything.
5. There may no longer be a macOS team, but there certainly still are engineers (and even smallish sub-teams) that are predominantly working on one platform or another.
And some of the differentiation that IS being done is not A/B testing in the usual sense of the word. i.e. if one particular hardware model needs a different version of an asset, those models can be assigned to a separate set, but that does not really make it an A/B test.