PaperFlow
GitHub Add to Chrome

BROWSER-NATIVE RESEARCH WORKSPACE

Research shouldn't be
split across tabs.

Read, annotate, ask, remember, and organize papers without leaving the browser.

Add to Chrome View on GitHub
arxiv.org/pdf/1706.03762 PaperFlow active
PaperFlow Reader showing Attention Is All You Need with page thumbnails, annotation tools, and page-aware AI

Read where you find papers. Think where you read them.

BEFORE PAPERFLOW

Research starts
in pieces.

A PDF in one tab. Notes somewhere else. A chatbot with no page context. A library that cannot remember the conversation.

FRAGMENTED
PDF1706.03762
Attention Is All You Need
NOTESdraft.md

multi-head attention

Compare representation subspaces

SEPARATE AI TABoutside paper

What is this paper about?

AI RESPONSE

It introduces the Transformer, an architecture built entirely on attention. Self-attention lets each token relate directly to every other token.

General summary · no page or citation link
LIBRARY126 items
TitleStatus

Attention Is All You NeedReading

BERTSaved

12 browser tabs
PaperFlow ONE PAPER · ONE CONTEXT
PaperFlow consolidates the PDF, annotations, and AI into one workspace
CONNECTED ONE WORKSPACE

PAPER WORKSPACE

One paper.
One persistent workspace.

The source is not an attachment to the work. It is where the work lives.

PAPER / 1706.03762 Vaswani et al. · 2017
Attention Is All You Need
PAGE 03 / 15
READERPage 3 in view03 / 15
NOTES2 annotationssaved locally
AI3 paper conversationspage-aware
MEMORY1 saved insightpersistent
STATEReading positionsynced
METADATAREADING STATECONTEXT

SOURCE-LINKED AI

AI that stays
connected to the paper.

Questions, answers, and citations remain grounded in the page you are reading.

PaperFlow Reader with page-aware AI beside the source
✦Ask PaperFlow
Why is self-attention useful here?
PaperFlow

Self-attention gives each token a direct path to every other token, preserving long-range context without recurrence.

ANNOTATIONS

Think directly
on the paper.

Highlight an argument. Underline a definition. Attach a note to the exact place that made you stop.

PaperFlow annotation controls on a PDF
HIGHLIGHTUNDERLINENOTEAREA

PAPER MEMORY

Come back days later.
PaperFlow remembers where you left off.

ATTENTION IS ALL YOU NEEDPAPER MEMORY / LOCAL
READING STATE Page 3 · Scaled Dot-Product Attention 63%
PREVIOUS QUESTION Why is self-attention useful here? THREAD 03
SAVED NOTE Compare representation subspaces across attention heads. PAGE 03
CONTEXT 1 saved insight · 2 annotations · 3 conversations READY
NO RECONSTRUCTION

The paper opens with its context intact.

RESEARCH LIBRARY

Your papers become
a research system.

Collections, tags, status, citations, and history stay dense enough to scan and quiet enough to think.

PaperFlow research library with collections, searchable metadata, and reading status

LOCAL-FIRST ARCHITECTURE

Your research
stays yours.

PaperFlow keeps the document workspace in the browser. Sync and AI remain explicit connections you control.

LOCAL-FIRST NO DOCUMENT BACKEND ENCRYPTED SYNC OPEN SOURCE

PRODUCT PHILOSOPHY

Research tools should
disappear into the work.
NOT ANOTHER CHAT WINDOW

The PDF remains central. Everything else works beside it.

PaperFlow AI

Read where you find papers.
Think where you read them.

Add to Chrome GitHub ↗

OPEN SOURCELOCAL-FIRST