At the time of the Revolution, Cuba was effectively run by American East Coast mobsters and U.S. sugar, fruit, and tobacco interests. Security services like the relatively-new CIA got much more interested after it "fell" to communism, but were also part of the pre-Revolution power structure too -- as were the well-heeled Cuban oligarchs/capitalists/landowners who were dispossessed during the Revolution and decided to flee to Miami (and eventually produce our current Secretary of State)

I think something similar could be true today, and it doesn't require any natural resources beyond cheap labor, Caribbean weather, and an obedient government.

To put it another way: a closed-weight model is, by definition, impossible to independently benchmark.

Sure you can, just do it silently and don't tell the people hitting your API that the model is different now. Unless it's open weight, we're just taking your word for it. Even better, do a VW and try to detect which benchmark is running, then change to a hyper specialized model that is trained on it.

In engineering of all kinds (or at least the ones I'm familiar with), nothing really beats calmly sitting with your thoughts, stating a problem, then getting up and walking around while you think about it, then sitting back down to write down a possible solution, and then asking colleagues to read it.

The Highland Park plant was Ford's play to cut the Dodge Brothers out of the process by machining most of his own parts. The peak of vertical integration would be the River Rouge plant which, as you say, machined all its own parts from iron and steel made on-site from raw ore (but never made the Model T).

I remember calling the library reference desk from the phone behind a bar to settle bar bets. Once free long distance became a thing, you could justify calling west coast libraries during east coast happy hour, and I had a hand-written list of phone numbers on a piece of folded up yellow legal pad. The LA Central Library seemed to be the most patient with drunk midwestern college students shouting questions about medieval art at them. Now bets are settled before they even really get going, so it doesn't even feel fun to bet on, so people don't.

I've also taken several trips to Europe and only on the most recent one did it make financial sense for me to get a local data plan. I admit that the language of the country we visit is kind of a hobby of mine, and so talking to the locals is a lot of the fun of going, but even if that's not the case, what's wrong with a little mystery? You can snap the photo, and then for years down the rode if you show it to somebody, you can say "Here's a cool statue I saw in Serbia, but I'll be damned if I can tell you what the inscription on the plinth says." Or even 3 years ago, you probably would have posted it to $SOCIAL_MEDIA_PLATFORM with a caption like "Who can tell me what this says?" and perhaps even gotten a reply from somebody in the same city you were in and made a little connection.

It remains to be seen if LLMs will be used in this way, but the messaging so far (including your post, the one I'm replying to) suggests that the elites' inclination is exactly that: more profit, more control, more power now for elites, temporary (possibly years, maybe as long a generation or three) but real suffering for many of the rest, and a steady state at the end of the transition where they are still in charge, much richer, and maybe also everyone else is better off (if they can't capture the gains completely for themselves).