I don't understand how you can scale the connections physically. It is easy when you have just a few cells, but how would you even do it with 3000?

I would assume the goal is to keep expanding these circuits. If you compare initial microprocessors with today's devices with billions of transistors, even a fraction of a similar scaling would make them quite powerful.

how is thermodynamic linear algebra different from monte carlo algorithms? I'm willing to believe that there are efficiency gains to be made by exploring the time/energy trade-off, but i bet they have a long long way to go before they are better than general purpose compute (CPU/GPU) running well designed monte-carlo algos...

I think the benefit here is that noise is not something you need to calculate with eg a seed, but rather the device can focus on the actual computations required.

> how is thermodynamic linear algebra different from monte carlo algorithms

You may think of it as case of hardware for generate ideal noise on steroids.

And in typical computing system it is serious enough problem, that in some first Xeons included hardware noise generator.

https://en.wikipedia.org/wiki/RDRAND