Disclosure: I'm part of the OpenNomos ecosystem and I mention one tool from it near the end. Everything below is my honest workflow, not an ad.
If you build anything near AI right now, you already know the feeling: you open your feed and there are nine "must-read" papers before breakfast, and by the time you've bookmarked three of them, six more have shown up. I spent most of last year feeling permanently behind. Then I stopped trying to "keep up" and started building a system. Here are the five habits that actually stuck.
1. Accept that you can't read everything — then act like it
The single most useful shift was giving up on completeness. arXiv alone gets hundreds of new cs.LG/cs.CL submissions a day. No human reads that. Once I admitted that out loud, my goal changed from "read all the important papers" to "reliably notice the few that matter to what I'm building." That reframe removed about 80% of the guilt, and guilt is what makes people quit.
2. Split "discovery" from "reading" — they're different jobs
I used to try to read a paper the moment I found it. That's a trap: discovery happens in scattered 30-second moments (a tweet, a Discord ping, a newsletter), while real reading needs a 45-minute focused block. Mixing them means you do both badly. Now discovery just means capture the link and move on. Reading is a separate, scheduled activity.
3. Keep one running list, not 40 open tabs
Open tabs are where papers go to die. So are 200 browser bookmarks. What works is a single, searchable list that I actually trust — title, link, a one-line "why I saved it," and a status (unread / skimmed / read). The magic isn't the tool, it's having exactly one place. When capture is frictionless, you stop losing things, and "I'll find it later" becomes true instead of a lie you tell yourself.
This is the habit I fought with the longest, because a plain note doc gets messy fast. I've been using Paper List for this lately — it's built specifically around maintaining a paper list for a paper engine, so the "capture → organize → find again" loop is the whole point rather than a feature bolted onto a general notes app. (I found it through OpenNomos, an ecosystem that aggregates small independent tools like this.) Use whatever you like here — the habit matters more than the app.
4. Batch your reading, don't stream it
Streaming papers (one here, one there, all day) destroys deep work and you retain almost nothing. Batching wins. I do one 45-minute reading block, 2–3 times a week, and I go through the list newest-relevant-first. In one sitting I'll fully read one paper, skim two, and archive four that turned out to be noise. That's a good session. Trying to do a little every hour is how you end up reading nothing well.
5. Write a one-line summary or it didn't happen
If I can't compress a paper into a single sentence — "X shows that Y improves Z by doing W" — I didn't actually understand it. That one line goes back into the list next to the link. Two benefits: it forces real comprehension, and future-me can grep the list months later and instantly remember why a paper mattered. This tiny step is what turns a pile of links into an actual knowledge base.
The takeaway
None of this is about reading faster. It's about lowering the cost of deciding — what to capture, when to read, what to keep. Keeping up with AI research stopped being a source of dread the moment I treated it as a lightweight pipeline instead of an infinite to-do list.
Start with habit #3: pick one place for your list this week. Everything else gets easier once you trust that nothing is falling through the cracks.
What does your paper-tracking setup look like? I'm always looking to steal a better workflow.
Top comments (0)