
How are you keeping up with ML papers without letting arXiv, social feeds, or newsletters decide the whole reading list?
I am trying to understand how ML researchers handle the paper firehose when they are not doing a targeted search.
Search works when you know the question. The harder workflow seems to be the lighter scan: which topics, authors, labs, journals/sources, or adjacent areas should stay on your radar?
I built Scollr around that second workflow. It is my product, so I am not pretending to be neutral, but the use case is pretty specific: follow research topics/authors/sources and scan a feed of papers that may be worth opening.
For people who read ML papers regularly, what would make that kind of feed useful or useless? Is the main problem coverage, ranking, freshness, topic control, or trust?