This is the source for the leap second predictor that is hosted on https://www.565851109.xyz - it also includes some code for manipulating Bulletins that may be useful
Jenkins takes care of doing this automatically and uploading to S3, but manually, using Poetry:
poetry install
mkdir dist
poetry run bulletin-aThis repo needs to be updated weekly when the new IERS Bulletin is published. This repo includes a script to automatically fetch the files and update this repo using the Gitea API, but it is pretty specific to my setup and probably isn't much use for others.
The math is a bit beyond me, but the bulletin contains a formula like:
UT1-UTC = dut1 +/- rate * (MJD - mjd1) - (UT2-UT1)
and
t = 0.00025 * (MJD-mjd2) * 0.75
where dut1, rate, mjd1 and mjd2 are constants, and MJD is the date to predict for.
Borrowing Tonys' code, you can calculate a predicted value for UT1-UTC for an arbitary date in the future. I'm picking the 1st of July and 1st of January to see UT1-UTC is greater than 0.5 or less than -0.5, in which case a leap second would be needed. If UT1-UTC is negative, then a positive leap second would be needed, otherwise a negative one. I find that trying to calculate predictions that are too close to when the bulletin was issued produce nonsense answers, and I stop calculating if there is no answer by 2060.