Hi Product Hunt! Attention is the scarce resource: compute scales freely, a researcher's reading day does not. I don't trust a tool that reads the papers for me and hands back a verdict; that takes the one step I won't delegate. So Skim doesn't move judgment off your plate, it moves the read step off the critical path. It pre-walks the queue while you work and marks the lines that warrant a pause, in the open, with the call left to you. That's the problem and it isn't summarizing the paper in front of you. Pasting a PDF into GPT or Claude already works for the one you've decided to read. The cost is the 100+ a week you'll never get to: the queue, not the PDF. So Skim triages the queue, it doesn't summarize a paper. It walks your reading list while you work and scores every line against a small reader profile you keep in YAML. A line gets tinted in place, written to your notes (Obsidian / Notion / Markdown, deep-linked to PDF#page+line so checking it is one click), or rarely, ~2×/week earns an audio (mail) alert. You don't need an invite to judge it: the demo runs live on the site right now. What it's NOT: a benchmark-chaser or a one-paper summarizer; where it's weak is written down (cold start, heuristic scorer): https://paperskim.org/why Invite-only; request during launch day and you skip the line into this week's batch. What I'd genuinely like feedback on: 1. Is triage-over-the-queue a real pain for you, or a nice-to-have? 2. Where would captures need to land to fit your workflow? 3. What would make the rare ~2×/week alert worth trusting? Live demo + request: https://paperskim.org and I'll read every reply here. One more thing for the architecture-minded: the session log is the interface: reading, triage, and execution decouple, so the log composes with autonomous agents downstream. The triage stays transparent; the judgment stays yours.

Skim

A second pair of eyes on the papers you'll never reach

A second pair of eyes on the papers you'll never reach