For medicine
A two-column paper, a 300-page clinical guideline or a whole textbook reads like a book: one column, your type size, and tables as tables. The AI answers with the text of the document, not with what it remembers: every answer brings the passage and the page, and one tap takes you there. No account, no install, no upload.
Reading asks for nothing. An account is asked for only when you ask a question; it is free and only needs an email.
GBD 2019 Diseases and Injuries Collaborators · Lancet 2020; 396
| Age group | DALYs 2019, count (millions) | Age-standardised rate (per 100 000) | Annualised change, 1990–2010 |
|---|---|---|---|
| 0–9 years | 531 (458 to 621) | 19125·7 (16495·1 to 22382·5) | −2·5% (−2·6 to −2·3) |
| 10–24 years | 229 (194 to 270) | 12313·0 (10399·9 to 14478·3) | −0·7% (−0·8 to −0·6) |
| 50–74 years | 832 (752 to 919) | 28263·2 (25527·6 to 31213·4) | −1·0% (−1·0 to −0·9) |
| ≥75 years | 329 (308 to 351) | 77320·5 (72372·5 to 82440·3) | −0·9% (−0·9 to −0·9) |
Between 1990 and 2019, the number of global DALYs remained almost constant, but once the effects of population growth and ageing were removed by converting counts to age-standardised rates, there were clear improvements in overall health. Over the past decade, the pace of decline in global age-standardised DALY rates accelerated in age groups younger than 50 years compared with the 1990–2010 time period (table). The annualised rate of decline was greatest in the 0–9-year age group.
The burden for children younger than 10 years declined profoundly between 1990 and 2019, by 57·5% (95% UI 50·3–63·1). Key drivers of this progress included large reductions in major infec-
tious diseases affecting children, namely lower respiratory infections, diarrhoeal diseases, and meningitis, each of which declined by more than 60% between 1990 and 2019. In 2019, neonatal disorders were the leading cause of burden in this age group.
1209
What you see above: pages 5 and 6 of that paper, as they come out of the converter. Two columns become one, split words are joined, and the table stops being a picture: it reads at any size, it can be selected, and the AI understands it row by row. The text is trimmed to fit; nothing else is.
The original page is still there. Switch whenever you want.
What it is for
It is not only for papers. What changes between a 12-page article, a 300-page report and a 1,200-page textbook is the table of contents and the search; everything else works the same.
Try it: press a marker in the answer.
Where did the burden fall the most between 1990 and 2019?
In children younger than 10, whose burden declined by 57·5% (95% UI 50·3–63·1). The paper attributes most of that progress to large reductions in lower respiratory infections, diarrhoeal diseases and meningitis, each down by more than 60%.
“The burden for children younger than 10 years declined profoundly between 1990 and 2019, by 57·5% (95% UI 50·3–63·1).”
“Key drivers of this progress included large reductions in major infectious diseases affecting children, namely lower respiratory infections, diarrhoeal diseases, and meningitis, each of which declined by more than 60% between 1990 and 2019.”
The burden for children younger than 10 years declined profoundly between 1990 and 2019, by 57·5% (95% UI 50·3–63·1). Key drivers of this progress included large reductions in major infectious diseases affecting children, namely lower respiratory infections, diarrhoeal diseases, and meningitis, each of which declined by more than 60% between 1990 and 2019.
What it does not do
With this audience, overpromising is expensive. This is what Marginalia is not and what it does not do well yet with papers and guidelines.
Privacy
What it is tested with
The converter is tested with open-access papers from these journals, and the reading order is compared with the text published by PubMed Central. All above 97% agreement, with tables rebuilt as tables.
If your PDF comes from another layout and comes out wrong, send it: every new layout goes into the tests.
Questions
Trust the citation, not the answer. The AI answers with the document in front of it and can still be wrong; that is why every claim carries the passage and the page, and one tap opens the text as written. If something is not in the document, it says so.
Yes. It is converted once, in your browser, and kept; the table of contents comes from the PDF and search covers the whole document. On phones with little memory the first conversion of a very large PDF can take a while.
Only if they have a text layer (a decent OCR). If the PDF is an image, it reads in its original view but cannot be reflowed or asked about. It tells you when you open it.
To read it, yes: the PDF does not leave the browser. When you ask, the question travels with the passages needed to answer it, so with a document carrying patient data it is better not to ask, or to anonymise it first.
It is good for studying from the sources: textbooks, guidelines and papers, with highlights, notes and questions to the text. It is not a question bank or an exam simulator.
Nothing. Reading needs no account; asking does, and it is free with a daily and monthly cap that covers an afternoon of studying with several papers. If that ever changes, it will be said here first.
Start
No account and nothing to install. The sample paper converts in a few seconds; yours, as long as it takes to open. If you read often, it installs as an app.