Just talked to a customer today that we developed an application for 10 or so years ago. We didn't win the follow-on builds for political reasons. What I learned was that our system and the newer versions' learnings were scrapped in favor of "doing it better our way". Well, the first systems used FPGA/embedded devices. The current vendor is trying to replicate the embedded devices with the equivalent of a 386 processor running DOS. The application was precision load control running at 50kHz sample rate with a load fault detection time of less than 100us. Not very possible with non-realtime systems. Customers will always want to have an application "their way" because Burger King said they could have it.

Not to hijack the thread further, but I believe this is quite common. I was working on a government contract where we were required to develop two separate sets of equipment to support two assembly processes for two different customers of the same contract. There were two teams running in parallel and there were not supposed to share information between the teams even though a lot of the team members worked for both teams. When election season was over, all the top brass changed which changed the scope of the project. We were not allowed to discuss which customer was which and only refer to each process by the project code name. The development times were decadal, so senior engineers would retire and come back several years later to pick up where the last activity was cancelled and mothballed by the government.

Great questions. I have recently started using Python for machine vision prototyping as well as exploring the machine learning libraries. While I like the language, the lack of structured architecture is a bit unnerving. It reminds me of National Instruments LabVIEW which is also very easy to use and make a big mess in. I have had conversations with our software department and there are some very good libraries available in Python for complex math operations that are well documented and have community support. There are similar libraries available in C#; however, they are difficult to implement and lack proper documentation or support. I think this may have more to do with the underlying code that supports Python is C++ rather than C#, so there is some conversion/wrapping that needs to be done to make the library available to C#. I think it comes down to your use case, but I don't think there is anything available in Python that isn't available in C#. A lot of the available Python libraries are wrappers for other general purpose code blobs available in other languages. Case in point is the Kivy library for building UI interfaces. The code base underpinning Kivy is god-level genius. Some of the implemenation requirements are a bit clunky, but I have been impressed so far with the library.

Well, I have seen more out of Musk that makes me believe he understands social interaction better then Bill Gates. Bill Gates NEVER polled his user base about anything, just pushed his latest "improvement" to a captive audience. Don't even start about Linux on the desktop... Not that anyone is comparing directly, but Gates is a real oligarch and fascist. Musk seems to be more of an attention whore much like Trump.

Yes. LG has a couple models; however, they are mostly carrier locked and marketed to seniors.