about
Keep up with mathematics in a few minutes
Paperlane collects the new mathematics papers every weekday and sets them in one list that you can make your own in seconds. It is free and needs no account.
How search works
Type a topic or an author's name, and Paperlane ranks the papers with BM25, the standard formula behind most search engines. It reads titles, abstracts and author names, and it counts words in a title twice. A phrase that appears word for word gets a bonus, and British and American spellings match each other. Only papers that match are shown, best first, with the words that matched under each one. This runs in your browser and nothing you type is sent anywhere.
With no search, papers sit under their main field, largest field first, and the fields you picked move to the top. You can also list them newest first or shortest first. Paperlane never ranks a paper by who wrote it.
What Paperlane does for your AI
One night of mathematics is about 57,000 words of abstracts, the length of a short novel. Paperlane reads it once and hands your assistant an index of about 7,500 words, so the answer is fast and cheap. Some answers also need real work, such as checking every linked repository for a Lean project, and Paperlane does that once for everyone. Each issue stays in the archive for three months, so you can look back and see when a paper was revised.
What the tags mean
- ∎ lean
- A linked GitHub repository contains a Lean project file (
lean-toolchainor alakefile). It has not been built yet. - ∎ lean builds
- Paperlane cloned the repository, installed the Lean version it asks for and ran
lake buildon a clean machine, and the build passed with nosorry. The paper's page names the commit and the Lean version. This shows the project compiles. The project's page also lists any axioms its theorems use beyond the standard three. It does not show that the Lean statements match the theorems in the paper. - ∎ lean 3 sorry
- The build passed, but Lean reported declarations proved with
sorry, which means some proofs are missing. - ∎ lean fails
- The repository has a Lean project, but
lake buildfailed or ran for more than an hour. - □ lean
- The paper says it is formalised in Lean, but no Lean repository was found.
- code
- The abstract or comments link to a code host such as GitHub or GitLab.
- v2
- A revised version of an older paper.
- via cs.CC
- A paper from another subject that its authors also listed under mathematics.
Every tag comes from a simple check that anyone can repeat, and no language model decides any tag. Paperlane never scores authors.
Every checked Lean project has its own page with the commit, the Lean version and the papers that link it, and lean builds lists them all. Each page also has a badge that authors can put in their README.
When it updates
New papers are announced from Sunday to Thursday evening, US Eastern time. Paperlane picks them up at 06:28 UTC the next morning, so a new issue appears from Monday to Friday.
For agents and tools
Everything on Paperlane is also published as plain files. Point your assistant at any of these.
| llms.txt | A guide to every file, written for language models |
| latest-titles.md | One line per paper: id, fields, tags, title, authors |
| latest.md | The whole day with abstracts |
| latest.json | The whole day with every field |
| latest/math.CO.md | One field of the day (swap in any field code) |
| p/2611.21654.md | One paper (swap in any arXiv id) |
| feed.xml | RSS for your feed reader |
| lean/papers.json | Every Lean build result, keyed by arXiv id |
Privacy
Paperlane counts visits with GoatCounter, which sets no cookies and stores no personal data. It also counts a few anonymous events, such as a prompt being copied or a Markdown file being opened, to learn which features people use. Your fields and interests are kept in your own browser and are never sent anywhere.
If you sign up for the email, your address is kept by Kit, which sends it. Kit counts opens and clicks, as most email services do. Paperlane never shares your address.
Where the data comes from
Titles, authors, abstracts and categories come from arXiv, and every paper links to arXiv for the full text. Papers that Paperlane has not listed are looked up through Semantic Scholar or OpenAlex. Thank you to arXiv for use of its open access interoperability.
Paperlane is open source on GitHub, where you can report a problem or ask for a feature.