They kind of did explain it. If they cashed it in then there would be coverage about who won it and the fact that they were American and were placing bets with a system would have surfaced. So basically they could make more money by continuing to slowly siphon money off then they could with a big win like that.
Not the person who commented earlier but it feels a bit odd to mix the settlement payments and the actual spending transactions in the same list. Not sure if a different view would be useful or whether it is just something I would get used to over time.
You can use Intel PlaidML right now on OSX to utilize your metal supported GPU for machine learning. I have tried the python library with Keras and my GPU (Geforce 760) with no issues.
I used syncthing for a while but I found it was murdering my CPU. Didn't bother trying to solve that problem - just uninstalled it since the value it was providing me was less than any troubleshooting effort would have been.
I am not a homeless expert but I think a non-insignificant portion of homeless are not mentally stable enough to accept help and are suspicious of anyone that attempts to help them get off the streets.