Digest and remember properly the last discovery and work about your favorites topics
How it works
Three steps. Sign up once. The rest takes care of itself.
We scan your arXiv categories every morning and surface the one paper worth your time — no feed, no bookmark pile.
The paper deep-read across four angles — what they did, what they claim, what to trust, what the counter-argument is.
A short comprehension loop catches the part you skimmed — so you actually retain the paper, not just read it.
No card · 3 deep-dives/month free · cancel anytime
Not a feed. Not a summary. A deep-read.
We don't dump a feed of fifty links. One paper a day from your categories — the one worth your coffee window.
Methods, results, limitations, implications — so you walk away knowing what the paper actually claims, not just what the abstract says.
Every claim is anchored to a verbatim quote from the paper. No hallucinated summaries. No vibes-based reading.
A short Socratic loop after the read catches the part you skimmed. Five minutes, no grade, no shame.
Comprehension check now ships with every digest — see what you actually retained.
See changelog →From cs.LG to math.ST to astro-ph — every category arXiv supports.
From the founder
Antoine Pedretti
Founder, DIGEST
The day I gave up keeping up with AI research was sometime in December. I'd been trying to stay current the honest way: read the papers, not the threads. By the time I gave up, my open-tabs folder had three figures of arXiv links in it. The queue was growing faster than I could read it. I was further behind than when I'd started.
So I built the tool I needed. Then I turned it into a product. The bet: the bottleneck isn't discovery — it's the deep-read after. One paper a day. The angles that matter. A short comprehension check at the bottom so you know what stuck.
The morning I'd been dreading became a fifteen-minute coffee-window read of one paper, written for me. That's it. That's the whole product.
Notes on technique, paper highlights, and what we're shipping.
On the roadmap
Soon: query your DIGEST history from Claude, Cursor, or any MCP-aware client. “What did I read about diffusion samplers last month?” — answered from your own deep-reads, not the open web.
Free for a taste. Pro unlocks unlimited recipes + teach. Pro+ is for the power users.
Free
$0
Pro
$3/month
Pro+
$9/month
Feeds give you 50 papers to triage. DIGEST gives you one — already deep-read. The bottleneck is post-discovery, not discovery.
Every claim is anchored to a verbatim quote pulled from the PDF. If the paper doesn't say it, we don't either. The model frames; the paper speaks.
All of them — 150+ categories across CS, math, physics, biology, economics, statistics, and more.
3 deep-dives per month, one recipe, all six reader profiles. No card.
A short Socratic loop at the bottom of each digest — four checkpoints (summary, contribution, evidence, critique). Free: 1/week. Pro: unlimited.
Yes — full refund within the first 14 days, no questions asked. Email hello@digest.ltd and we issue it the same day.
Yes. One click in settings. No emails to write, no retention call.
Fifteen minutes for the full deep-read plus the comprehension check. The whole point is a coffee-window read, not a half-day commitment.
The picker only chooses from the arXiv categories you opt into. If nothing in your categories meets the bar that day, we skip — you'd rather get nothing than noise.
One email per send. No drip campaigns, no upsell sequences, no newsletter cross-promo. The only email outside your digest is account-critical (password reset, billing).
Yes — pause anytime in settings. Billing pauses with you and resumes when you do. No tier downgrade, no data loss.
Every digest has a shareable link to the per-section landing pages — evidence, perspectives, teach. They can read without an account. If they want their own picks, they sign up.
Not yet. v2 is arXiv-only by design — it covers 150+ categories and is where most preprints land first. Multi-source ingestion is on the roadmap.
Your digest stays anchored to the version we read. The link always resolves to the latest version on arXiv, so you can spot what changed.
