Cinepsus
An LLM reads every page of your paper alongside you.
Start readingEvery answer cites the page it came from. Dense passages explained on selection. And sandboxed Python to reshape the paper's figures your way.
Your library. Drop a PDF, and it's ready before you are.
Reading.
Upload a PDF and Cinepsus reads the whole paper in the background while you start on page one. Ask anything in the chat beside the paper — answers stream in grounded in the full text, each with a citation pointing to the exact page it came from. Scanned papers included.
The paper and the conversation, side by side — with page chips linking every claim back to the source.
Understanding.
Select a passage, box a figure, or highlight an equation. Cinepsus explains it in place — LaTeX-rendered walkthroughs for the math, chart understanding for the visuals.
Highlight the equation — get the walkthrough in the paper's own notation.
Box any figure — the trend, the axes, and what the authors want you to see.
Select any passage for an instant, grounded explanation — Explain, Simplify, or Why it matters.
Step-by-step derivations, rendered in LaTeX, from the notation the paper actually uses.
Box a plot and get the trend, the axes, and what the authors want you to see.
The paper's data, your figure. Replot a chart, restyle a comparison, or sketch a custom visual — sandboxed Python, zero setup.
Grounding.
Every
page
read before you finish the abstract
Every answer
cited
to the page and section it came from
Every source
labeled
full read, partial, or abstract-only — you always know what grounds an answer
Comparative studies.
This is where Cinepsus earns its name. Pick the papers, and it reads them all, links methods, datasets, and claims into one knowledge graph, and grounds a single chat across the whole set. Ask "whose method wins on sparse data?" and get one answer, cited paper by paper. Referenced works join the study at whatever depth exists — full read, partial, or abstract-only, always labeled.
One study, five papers — methods, datasets, and claims linked across the whole set.
See how concepts connect across the study, and jump from a node straight into the chat.
Ask one question across all the papers and get a comparison, cited paper by paper.
Every conversation kept per paper and per study — with titles the AI names for you.
Cinepsus reads the full text of your paper before answering. Every response carries a citation to the page and section it came from — click it to jump there in the PDF.
Yes. Scanned PDFs are OCR'd during ingestion, so highlighting, chat, and citations work the same as on born-digital papers.
It resolves every reference it can: some come back as a full read, others partial or abstract-only, and a few are unavailable. Each one is labeled with its depth, so you always know exactly what grounds a comparison.
Small, sandboxed Python runs over the paper's own data — restyle a figure, replot a comparison your way, or build a custom visual the authors didn't include. It's not for replicating the whole study; it's for seeing the paper more clearly.
A workspace holding several papers with one shared chat, a knowledge graph across them, and AI-named conversations — built for literature comparison, not just single-paper reading.