On April 15, the National Institute of Standards and Technology ( NIST ) in Boulder, Colorado, claimed, in the journal Metrologia, to have developed the world's most accurate timepiece, to less than 1 sec/100 million years.

Here is the article Accuracy evaluation of primary frequency standard NIST-F4 - IOPscience

that details the NIST-F4's workings.

This would mean that having synchronized your watch with a T-Rex 100 000 000 years ago, it would be currently off by less than a second.

I'm not sure how 'ground-breaking' this is, as I'm not aware of the accuracy of current atomic clocks, and most of us are fairly ignorant of such things.
Perhaps @swansont can elaborate further and educate us ...

The first thing I’ll note is that it’s a frequency standard, not a clock, but that’s a pet peeve that basically only bothers me. (clocks run continuously)

I’m guessing that the new cavities they mention incorporates a lot of cavity design modeling that Kurt Gibble has done over the years. The basic fountain design is decades-old technology by now, but the best you can do for a primary standard, which has to use cesium and probe the microwave hyperfine transition (a spin flip of the ground states)

By tossing the atoms up at a few m/s you get them to be in a superposition of those states for a good fraction of a second, so it’s billions of oscillations, and since it uses the same cavity you eliminate certain errors you get with a two-cavity beam apparatus. By making the atoms cold you can get some fraction to come back down on-axis; the thermal speed at the temperatures they run at is of order 1 cm/s, so the ball of atoms they toss slowly expands.

Since it’s a primary standard, much of the paper is the assessment of the various frequency shifts and uncertainties in them that are present. (Trivia questions that imply that you’re probing at 9,192,631,770 Hz are wrong; there’s a magnetic field present that moves you away from that, but you can determine that field so you know what the shift is. Not being at 0 K also gives you a shift) That requires a lot of careful work

It’s an incremental improvement more than being ground-breaking, much like shaving a millisecond off a world record sprint time. It’s still a new record.

Thanks.
I saw the claim of it being the most accurate 'clock' ever, and I thought I'd ask an expert, as understanding the technical article would have required a lot of research on my part.

So what is a frequency standard used for ?
Calibrating atomic clocks ?

And if this new standard has only an incremental accuracy improvement over standard atomic clocks, wouldn't calibration with such a standard provide only minor improvements in accuracy ?

4 hours ago, MigL said:

Thanks.
I saw the claim of it being the most accurate 'clock' ever, and I thought I'd ask an expert, as understanding the technical article would have required a lot of research on my part.

So what is a frequency standard used for ?
Calibrating atomic clocks ?

Yes, basically. They realize the second, or a frequency tied to that, so you can tell if your clocks are running fast or slow, and make adjustments as necessary. So you’d have a bunch of clocks, and do the calibrations once in a while. It can be a year between these assessments; they sometimes dismantle the standard to modify it.

4 hours ago, MigL said:

And if this new standard has only an incremental accuracy improvement over standard atomic clocks, wouldn't calibration with such a standard provide only minor improvements in accuracy ?

Yup, but small improvements add up over time. Most people aren’t going to notice the result but it does push other standards labs to make similar improvements.