The idea is to have water or molten salt cool the walls of the torus from outside, and those drive ordinary turbines like any other generator. The main issue is that particles fly out of the confined plasma donut and degrade the walls, whose dust flys into the plasma and reduces the fusion efficiency. They’re focusing on the hard part - dealing with the health of plasma sustainment and the durability of the confinement walls over time. Hot thing that stays hot can boil water or salt to drive regular turbines, that’s not the main engineering challenge. I get your frustration where it feels from news coverage that they’re not focusing on the right stuff, but what you’ll likely eventually see is that the time between “we figured out how to durably confine a healthy plasma” will quickly turn into “we have a huge energy output” much like inventors puttered around with flight for hundreds of years until a sustained powered flight design, however crappy, finally worked. From that point, it was only 15 years until the first transatlantic flight.

Lol have you not used o1/o3? They show the inner monologue too. Fun little pretend detail to keep you entertained while the model takes 30 seconds to respond.

Uh…yes it’s about the same, maybe slightly dumber than o3. Which is already mediocre.

I see Tiktok in the news, never downloaded it. Does that make me special?

I don’t get it. Why would DeepSeek supposedly get more downloads from the app store? It’s not like it’s any more interesting to consumers than other LLM apps. It was just notable for purportedly being cheaper to train. Something seems fishy.