← All posts
This post is written in English.
Useful quantum computing needs special ingredients called magic states, and they're expensive. Estimates suggest they could take up to 95% of a quantum computer's capacity, and each one is consumed when used. Google's post on a new paper in Science reports progress on making them.
What cultivation does
Think of it as a filter. A noisy state is injected into a small region of the chip, coordinated measurements look for errors, flawed attempts are thrown away, and a purer state grows and is then adapted into a larger encoding. The technique was proposed in 2024.
On Google's Willow processor, the team reports:
- About 40 times fewer errors than states without cultivation.
- A record logical fidelity of 99.99% for T-states.
- Grafting a state from one type of error-correcting code onto another.
What it doesn't mean
Google says the method still has to be shown at a larger scale before it's realistic for practical quantum computing, and that significant technical challenges remain. It's a hardware demonstration, not a working application, and it won't change anyone's encryption this year.
My read
I'm not a quantum specialist, so I'd take this mostly as a signal of where the work has moved. The hard problems used to be keeping qubits alive at all. Now the discussion is about the cost of specific steps in error-corrected machines, which feels like engineering more than basic physics.
It's also a good reason to take the slow move toward post-quantum certificates seriously, even without a date on the calendar.
