As with any quantum computing news, I will wait for Scott Aaronson to tell me what to think about this.
https://www.nature.com/articles/s41586-025-09526-6
In the last sentence of the abstract you will find:
"These results ... indicate a viable path to practical quantum advantage."
And in the conclusions:
"Although the random circuits used in the dynamic learning demonstration remain a toy model for Hamiltonians that are of practical relevance, the scheme is readily applicable to real physical systems."
So the press release is a little over-hyped. But this is real progress nonetheless (assuming the results actually hold up).
[UPDATE] It should be noted that this is still a very long way away from cracking RSA. That requires quantum error correction, which this work doesn't address at all. This work is in a completely different regime of quantum computing, looking for practical applications that use a quantum computer to simulate a physical quantum system faster than a classical computer can. The hardware improvements that produced progress in this area might be applicable to QEC some day, this is not direct progress towards implementing Shor's algorithm at all. So your crypto is still safe for the time being.