Now that I am a PhD student, I realized that it was “my job” to keep up with ePrint. In the recent past, I used to subscribe to the now defunct ePrint bot on Twitter, and in the distant past, when I was working in a different field, I used to subscribe to arXiv’s daily listings. However, with the number of ePrints coming out each week and my newfound interest in actually keeping up with the field, neither approach seemed optimal.

Screenshot of the now defunct ePrint bot showing its last tweet on June 17th.

So, I looked around to see what other people were doing. My former colleagues in quantum computing seem to have embraced the community voting approach with SciRate, which now seems to also have job postings and apparently NSF sponsorship (!). The machine learning folks have long had their Arxiv Sanity Preserver which surprise, surprise uses machine learning magic ✨ to preserve your sanity when browsing the arXiv.1 And finally, the data science folks seem to have taken a radically different path with Connected Papers which looks at papers in the context of other papers and visualizes the whole thing as a graph, if you haven’t checked this out yet, please go give it a shot now, it is amazing!

These approaches are amazing but I still find myself yearning for that human curation that draws me towards content like tl;dr sec and Last Week in AWS. But, alas, I couldn’t find a similar curation for ePrint. Now, I could wait for someone to come along and create such a high-quality publication or I could make something really crappy, so crappy that it essentially forces them to give up and put out such a publication.

Introducing Ciphertext Compression

First, we need a name. I was initially thinking something straightforward like “Last Week on ePrint” drawing inspiration from the infamous “Last Week Tonight” and the previous cited amazing newsletter “Last Week in AWS”. But then I saw the pqc-forum drama about “2x ciphertext compression” and I just couldn’t resist. I love the name “Ciphertext Compression” it is cute, slightly funny, and speaks to me on many levels.

Now, we just need a format and we’re in business. If we want people to read this, we want to aim for something like 500 words, maybe 1000 words tops. Just the titles, authors, and ePrint identifiers from last week’s ePrint updates is over 1500 words. So we need to aggressively prune this list. This seems easy enough and the ePrint homepage even includes a link to the ePrints from last week, but the list isn’t very skimmable, it is just a list of titles and authors. So I made a more skimmable version, click on the arrows in the gallery below to see the difference.

The more skimmable version is live2 and open-source, so please give it a spin.

With all that out of the way, now I just need to skim ~80 ePrints and write a short blog post, it’s getting late and I am already over the 500 words, so that’ll be the next post. Thanks for sticking till the end. 💜

  1. The ML folks also seem to have the SciRate-like Papers with Code↩︎

  2. Right now, the live version only updates on Friday, Saturday, and Sunday (because those are the days I’ll likely look at it), but if people want it to update on all days, lemme know. ↩︎