Built by Antoine Pedretti at Helmo Solutions Ltd. Here's what we're trying to do, and what we won't.
Three chapters — the problem, what we built, and what we're trying to be.
Anyone reading arXiv seriously has the same problem: the firehose. Hundreds of new papers a day across a handful of categories you care about. Even just scanning titles takes most of an hour, and most of what you scan isn't relevant to what you're doing this week.
The shortcuts most people use are bad in their own ways. Twitter / X surfaces the loud papers, not the right ones. Newsletter aggregators write for a general audience and lose the technical depth a researcher needs. Saving titles to read later turns into a graveyard.
DIGEST started as a tool for one user — me. The shape: pick the arXiv categories you actually read, pick a reading profile (the lens you read in — Student, Researcher, Industry Pro, Curious Adult, Quick Scan — the summarizer tunes per profile), and get a digest in your inbox at the hour you choose. Daily or weekly.
The summaries link straight back to the arXiv abstract. No content is hidden behind us. If anything looks interesting, you read the paper.
Honest about scope. arXiv-only at launch. English only. No fabricated papers — every citation links to a real abstract. No engagement tracking, no analytics cookies, no selling your reading profile. The whole product is the digest in your inbox.
If you've ever felt guilty about the arXiv tab graveyard, this is for you.
Promises by negation — what we'll never build into DIGEST.
Every summary links to a real arXiv abstract. If DIGEST claims something about a paper, you can verify it in one click.
No analytics cookies, no time-on-page tracking, no third-party pixels in your emails. The only cookies we set are for the Supabase auth session and the Stripe checkout flow.
We don't sell, share, or syndicate your reading preferences. Your categories + profile are used to render your digests — that's it.
We're not optimising for time-on-page. Every digest opens with the paper, links straight to arXiv, and ends. You're done in five minutes.
