Read the new papers that actually matter, and see why they made the cut.
WILLOW Research Radar sweeps the new arXiv papers in your field every day, scores every one against what you actually care about, deep-reads the best from full text, and emails you a short digest. Not a black-box summary you have to trust, a digest that shows its work.
Get the daily radar See a sample digest
THE PROBLEM
Staying current is a losing race. Hundreds of new papers land every day in a single arXiv category, and the tools that promise to fix that mostly hand you a confident summary of findings the paper never made. In 2025 people logged over 146,000 hallucinated citations, and roughly half of retrieval-augmented answers turned out to be fabricated. So you are left with two bad options, drown in the firehose yourself, or trust an assistant that quietly makes things up. Neither is a way to run research.
HOW IT WORKS
WILLOW runs in two passes, the same economics a good analyst uses. First it does one deterministic fetch of the newest papers across the categories you pick (multi-agent systems, verification, NLP, security, databases, whatever your field is), then it triages every new abstract on a harsh one-to-ten scale against a research profile you define. Only the papers that clear your threshold get the expensive second pass, a full-text deep read with real analysis and a plain note on why it matters to you. Everything is deduplicated against a seen-ledger, so tomorrow's run only spends effort on genuinely new work. At the end it reconciles its own counts (fetched, triaged, deep-read), so an arXiv outage shows up as an outage, never as a quiet day you misread.
WHAT YOU GET
Every morning, one email: - the handful of papers that cleared your bar, scored, newest first, - a real deep read of the top hits (from the full text, not the abstract), with the analysis and the ideas-for-you spelled out, - the ones that did not make it, listed briefly so you can see what was skipped and why, - an honest count at the bottom, how many were fetched, triaged, and deep-read, so you always know the radar actually ran.
WHO IT'S FOR
Anyone who has to stay current in a fast-moving technical field and cannot afford to either miss the real work or trust a summary that lies. AI and ML teams, R&D groups, technical founders tracking a specific research area. If you have ever pasted an abstract into a chatbot and wondered whether the summary was real, this is for you.
PRICING (the ladder)
- Free / dogfood: the internal radar we run on our own development, keyless and shipping today. - Solo: one research profile, the daily digest to your inbox. - Team: multiple profiles, shared digest, delivery to email or Slack. - Verified tier: the citation-audit upgrade, where every claim in the digest is span-grounded back to its source with a receipt. This rides our deep-read and external-walk verifier, the same verification discipline that is the core of what we build.
IP is licensed, never assigned. The radar stays ours, the digest is yours.
THE PROOF (dogfood)
We use this on ourselves, every day. WILLOW Research Radar is the tool AYA runs on its own development, the research profile is our own current focus (multi-agent orchestration, verification, de-identification, context compression, text-to-SQL grounding, LLM security), and the digest lands in the founder's inbox. The product's first user is the company that builds it. We are not selling you a radar we do not run.
HONEST NOTE
Straight facts about where this is. The radar is built and it passes our quality gate, and it composes only capabilities we already had (the deterministic feed fetch, the triage and deep-read spans, the digest and delivery). What it does not yet have is a long public record of live daily runs, because that needs a running mesh, a funded model key, and email access wired up, and those are switches we have not thrown for an outside user yet. The deep read reads the real paper, but the span-grounded citation receipt is the verified tier, it is on the ladder, it is not the free digest. We would rather tell you that than pretend the first email is already the audited one.
Get the daily radar
*This page is a specification. The capability it describes is not built yet, and nothing here is a claim that it runs today.*