Reading shouldn't be
a bottleneck.
The tool we wished existed when we were drowning in PDFs. Reads them. Remembers them. Answers questions about them.
The problem
Researchers spend more time finding papers than reading them, and more time reading than writing.
Every researcher we talked to had the same three browser tabs open: a folder of unread PDFs, a half-written literature review, and a chat with a generic AI that confidently makes up citations. The PDFs never got read. The review never got finished. The AI never got grounded.
The bottleneck isn't information — it's attention. So we built a workspace that treats your library as a first-class object, not a backup folder.
What's different
Every answer cites a page. Every claim is auditable.
Three principles guide everything we ship:
- Grounded, not generative. The model only answers from yourcorpus. If a paper doesn't support a claim, we say so — never an invented citation.
- Comprehension, not chunking.Most “PDF chat” tools split a paper into 500-word slices and hope. We parse equations, tables, figures, and captions as first-class structured content.
- Workspace, not chatbot. Reading a paper, asking a question, drafting a review, mapping citations — these belong in one tool, not five.
A small team. A lot of papers.
Engineering & product
Built ML infrastructure for retrieval systems. Read too many transformer papers in graduate school; now reads them faster.
Document AI
Specialists in layout-aware parsing — formulas, tables, figures, multi-column flow. We obsess over the parts every other tool skips.
Research-led UX
Embedded with PhD students, postdocs, and competitive intel analysts. Every screen is shaped by someone who reads for a living.
Independent. Researcher-funded.
Privately held. Funded by researcher-investors. Not VC-pressured, not ad-supported, not training on your papers. Our incentives line up with yours — keep your library yours.
Try the workspace.
Bring your own library.
Free to start. No credit card. Cancel anytime — your library stays yours.