As noted in the article, it could be a very sharp transition from "largest factorization performed so far is 3x7=21" to "we can factor real-world RSA." If you

As noted in the article, it could be a very sharp transition from "largest factorization performed so far is 3x7=21" to "we can factor real-world RSA." If you want to make a classical computing analogy, it's like we're struggling to make transistors with more than a single 9 of reliability, and obviously you can't do complex computation with a circuit where every step gives garbage 1-in-10 times. Except it's not's obvious. 90% reliability could probably be made to work with silicon transistors with bog standard error correcting logic at the hardware level. Quantum error is a little bit more troublesome to work with, but there are also no known theoretical reasons error correction wouldn't work at existing error rates. We just need better algorithms, which might very well exist. Or, the next generation of chips would offer more 9's reliability, and even with existing error correction just a few more sigma in reliability would put us over the tipping point of being able to make reliable large-scale systems.