The December issue of Nature Astronomy is out (vol. 9), and:
Rehemtulla, N. Coughlin, M. W. Miller, A. A. et al. The automation of optical transient discovery and classification in Rubin-era time-domain astronomy p 1764 s41550-025-02720-6
There is simply no way a human can keep up with Vera Rubin. NO WAY. The camera produces a >3200 megapixel image, every minute, every clear night. That’s roughly a petabyte of data every 2 months. The next best option, then, is to have algorithms filter down the output; after one or two stages, then the stream might- might- be something manageable. And certain problems, like variable stars or slowly moving objects, are inherently ‘filterable’. A few other problems lend themselves to distributed ‘eyeballs’- assembling a corps of human volunteers, to perform visual recognition tasks at scale.
Rehemtulla et al. brief us on the issue in Nature. The journal has paywalled it (as is typical), but fortunately(?) there’s a preprint available:
https://arxiv.org/abs/2512.11959v1
Standard disclaimer: there’s a reason for a preprint, versus a ‘print’- the “pre-” indicates an earlier draft. In this case, the difference between this draft (the -v1 version, not ‘v0’) and the Nature Astronomy published paper is a difference of a few days. I’m going to assume the edits are minor.