Be honest about how "keeping up with the literature" actually goes. You start a new project full of resolve. You run a few searches, open fifteen tabs, bookmark a folder you name something optimistic like "to read." For a week or two you check in. Then a deadline lands, the folder goes quiet, and three months later you stumble onto a paper that's directly relevant, published back when you were still checking — and you feel that small, specific dread of having been out of date without knowing it.
This isn't a discipline problem. It's a design problem. Search is pull: nothing happens unless you remember to go pull it. And remembering, reliably, every week, for the length of a multi-year project, is not a thing humans are built to do. The fix isn't more willpower. It's to flip the direction — to make the new work come to you.
That's what Discovery Agents are. You tell an agent what to watch, and it goes to work on a schedule: scanning the literature, ranking what it finds against your field and the papers you already keep, and writing a one-line reason each result matters. Whether or not you open Folio that day, the watching happens.
An agent is a standing question, not a one-off search
A search answers "what exists right now?" An agent answers a better question, over and over: "what's new since I last looked, that I'd actually care about?" You define what you care about once; the answers accumulate.
That small shift changes how the whole thing feels. The firehose becomes a feed. You're no longer hunting — you're glancing at a short, ranked list and deciding. And because the agent remembers everything it's already shown you, you never rediscover the same paper for the fourth time, which is half of what makes manual searching so demoralizing.
A finding tells you why, not just what
Here's the part most "we'll email you new papers" tools get wrong: a list of titles is barely better than the firehose. The work you actually needed help with isn't finding papers — it's deciding which ones are worth opening. So every finding an agent surfaces leads with a single sentence, written for your field, on the specific contribution. You read that line and you know, in about two seconds, whether to go further.
Tap around a real one:
One sentence, written for your field, on the specific contribution — not "this looks interesting." You read this line and decide in seconds whether to go further.
A real finding, annotated. Tap any part of the card.
The note is the headline; everything else is there when the note earns a closer look. The abstract sits underneath in full — we used to clip it to two lines, which meant it routinely ended mid-sentence, so we stopped doing that. And when a finding connects to something already in your library, it says so, because new work means more when you can see how it lands against what you already know.
Each agent hunts differently
The thing people don't expect is that the agents have personalities — and the personality is functional, not decorative. A persona is a hunting style. Pick Vanguard and you get the freshest preprints pulled off the live web and checked often. Pick Archivist and you get landmark, heavily-cited work and nothing noisy. Forager goes after under-cited gems in adjacent corners; Curator keeps only the genuinely significant and throws back the rest. Choosing a persona quietly configures the agent — how far back it looks, how selective it is, how often it runs — and you can override any of it.
Try a few. Each one shows the way it actually hunts:
Stands watch over your core field.
The real personas from Folio. Pick one and the agent inherits its hunting style.
Run one agent or, on Chair, run several at once — one for your core question, one for the adjacent field you're trying to break into, one for the method everyone suddenly seems to be using. Each keeps its own feed, and a single view ties the threads together across all of them.
What you do with what it finds
A feed only helps if you're ruthless with it, and the agent is built to make that easy. When something's worth keeping, one click drops it into your library — where it stays marked, so you're never left wondering if you already saved it. Everything else you ignore, and ignoring is the system working, not failing.
When you want to think rather than triage, ask the agent directly: summarize the themes this week, or which of these cut against my thesis? It answers from what it's actually found, and the conversation is saved, so you can come back to it without starting over. And once a week — or once a day, if your field moves fast — it writes you a short briefing: the two or three themes, and which one or two papers to read first. If you'd rather listen than read, it'll narrate that briefing so you can catch up on a walk.
It watches; you decide
None of this is autopilot, and we've been careful to keep it that way. The agent does the watching and the first-pass judging. The decisions that actually matter — what to read closely, what to cite, what to dismiss — stay yours, and you can see and disagree with the reasoning behind every finding. The goal was never to read more. It was to be quietly confident that the work worth knowing will find its way to you, so your attention goes to the research instead of the search.
Folio's Discovery Agents watch your field and brief you on what's new — with notes, chat, and audio. Join the waitlist.