I’ve chewed on Gidney’s ‘Falling with Style’ paper.
I recommend reading it if you would like to understand Shor’s Algorithm.

I’m somewhat unclear if the following applies to Shor’s Algorithm in general, or just the modified version used for the experiment.
But I’ve come to understand that the algorithm is a recursive series of steps, structured such that it will eventually factor anything.
Like … it could take longer than the age of the universe for some numbers, but the algorithm will do the job if you got enough cycles to spare.

What we are looking for here is quantum supremacy, and once Gidney has explained this much, its obvious from the graph above that we are not seeing it. Pure random noise outperformed the quantum computer.

I guess the thing I’ve not absorbed yet is, why was the quantum computer expected to not work? I know it was much too complex a system, and internal noise would overwhelm any processing. Gidney described being amazed that the IBM quantum system even let him configure his experiment and run it. Why did it lose so completely to a random noise generator, as in how could you possibly get worse than random noise?

Okay! Now we’re finally getting somewhere. I was trying to figure out if you use the computer to get the dog interacts with the abacus or how, and it turns out: no! You can do the factorization with just the computer, just the dog, or just the abacus. In the dog case, he threw a ball until his dog barked 3 times, then 5 times, and claimed this demonstrates the factorization of 15, because 15 is 3*5. The criticism is that, because you choose the number to be factorized, and then do special operations that only work for that particular factorization, all processes that are fine-tuned to produce the correct factors are equally valid, including training your dog to bark a certain number of times.