Research,
compiled.
Type a topic. ArcXiv reads the arXiv literature and writes you a real lesson — prerequisites, chapters, equations, diagrams, and the papers that actually matter.
Try diffusion models · mean-field game theory · quantum error correction · why transformers work
A lesson, not a search result.
Every compile produces a structured reader you can actually finish in an evening — written from the papers, not around them.
Prerequisite map
Before page one, ArcXiv plans what you need to know first — and links those concepts so you can stub-in any gap without leaving the lesson.
Long-form chapters
Not flashcards. Real chapters with sections, intra-narrative continuity, equations rendered in LaTeX, and figures generated from the data.
A reading list that's been read
Every paper on the list comes with a one-line reason it's there, a difficulty rating, and the sections worth your time. No 80-paper dumps.
Skim · Standard · Deep
Toggle a chapter's register without leaving the page. Brisk summary, current text, or a textbook-grade expansion of every nuance.
Regenerate any section
Hit Regenerate on any section, drop an optional hint (“use a cooking analogy”, “go deeper on the math”), get a clean rewrite in place.
Built on real papers
Lessons are grounded in actual arXiv abstracts and metadata. When ArcXiv cites a paper it links to the canonical record at arxiv.org.
The arXiv interface that should have shipped in 2014.
Keyboard-driven feed, instant search, a personal library, and a paper-detail page that respects long-form reading. The whole corpus — every category, every preprint — at app-speed.
Three steps, no setup.
- Step 01
Type what you want to learn
One sentence is enough. “What's actually going on inside diffusion models.” ArcXiv plans the curriculum before it writes a word.
- Step 02
Read the chapters
A long-form reader with prereq chips you can click, equations in real LaTeX, figures rendered from the spec — built to be finished, not scrolled past.
- Step 03
Open the papers
Every cited paper lands in ArcXiv's reader. Save anything to your library, jump straight to the recommended sections, never lose the thread.
What is ArcXiv?
ArcXiv is two products glued together: a fast, modern reader for arxiv.org, and a learning compiler that turns any research topic into a structured, long-form lesson grounded in real papers.
The reader covers every top-level arXiv category — computer science, mathematics, physics, statistics, quantitative biology, quantitative finance, electrical engineering, and economics. The compiler reads the literature for you and produces a curriculum that's actually finishable: prerequisites first, chapters next, a reading list that someone has thought about, and a paper-detail page tuned for long-form reading instead of for archiving.
We don't host the papers — arXiv does — and we don't try to replace Google Scholar or Semantic Scholar. We're the layer on top: the thing that lets you go from I'm curious about this to I actually understand this in one evening.
Browse by category
Questions
How is this different from asking a chatbot?
ArcXiv's compiler plans a curriculum, picks canonical papers from arXiv, and writes a structured reader — chapters, equations, figures, a reading list — instead of a chat transcript. Sections cite the papers they came from and link to the arXiv records, so you can always verify and dive deeper.
What does a lesson actually look like?
A long-form, scrollable reader. Each lesson has a prerequisite map, several chapters broken into sections, real LaTeX equations, generated diagrams, optional self-check checkpoints, and a re-ranked reading list with a one-line reason per paper. You can toggle any chapter to Skim or Deep, regenerate any section with a hint, and pin concepts you want to revisit.
Is ArcXiv affiliated with arXiv.org?
No. ArcXiv is an independent project. arXiv.org is operated by Cornell University and remains the canonical archive — we read its metadata via the public APIs and link out to the original PDFs and abstracts.
Is ArcXiv free?
Browsing, search, the personal library, and the reader are free with no account required. A free account gets five lesson compiles per UTC day. Email hi@chaitanya.science if you need more.
Which research areas does ArcXiv cover?
Every top-level arXiv category: computer science, mathematics, physics, statistics, quantitative biology, quantitative finance, electrical engineering and systems science, and economics. If a paper is on arXiv, it's on ArcXiv.
How does ArcXiv stay accurate?
Lessons are scoped to arXiv. The compiler resolves concepts and papers against the corpus before writing, the section pass is conditioned on the actual paper metadata, and citations link to the canonical arXiv records so you can always verify.
Pick a topic. Read it tonight.
The compiler is free to try, no card. Five lessons per day.