DEVONthink
A document and research manager that finds related text by meaning, not just keywords.
Where DEVONthink comes up1 episode or test
- 1Load own sourcesOnly his own reading notes
- 2Ask open questionNot keywords, e.g. surprising dolphin facts
- 3Tool surfaces connectionsScans notes by meaning
- 4Human picks keepersJohnson chooses what to use
From Steven Johnson on DEVONthink
What was said about DEVONthink
DEVONthink surfaced a relevant quote that shared no keywords with the search. In the early 2000s it could tell two paragraphs were close in meaning even with no shared words. A search on waste recycling returned a quote about skeletons forming from calcium waste, which became about seven pages of The Ghost Map.
Steven Johnson, author of Where Good Ideas Come From, used it to search his own reading notes. The tool found the connection, and he chose to use it.
Search your own notes by meaning, then curate what the machine surfaces. Johnson's workflow is to load only your own sources and ask open-ended questions rather than keyword searches, such as the most surprising facts about dolphins in your reading notes. After writing a paragraph, he asks what else in his reading history is relevant.
He sees the human as curator at the start, the model scanning and surfacing connections, and the human picking the keepers at the end. This is the idea he carries from DEVONthink into AI tools like NotebookLM.
The Expert Behind Google’s NotebookLM AI Writing Tool
In the early 2000s it could tell two paragraphs were close in meaning with no shared words. A search on waste recycling surfaced a quote about calcium in skeletons, which became about seven pages of The Ghost Map.
DEVONthink surfaced a relevant quote that shared no keywords with the search. In the early 2000s it could tell two paragraphs were close in meaning even with no shared words. A search on waste recycling returned a quote about skeletons forming from calcium waste, which became about seven pages of The Ghost Map.
Steven Johnson, author of Where Good Ideas Come From, used it to search his own reading notes. The tool found the connection, and he chose to use it.
Search your own notes by meaning, then curate what the machine surfaces. Johnson's workflow is to load only your own sources and ask open-ended questions rather than keyword searches, such as the most surprising facts about dolphins in your reading notes. After writing a paragraph, he asks what else in his reading history is relevant.
He sees the human as curator at the start, the model scanning and surfacing connections, and the human picking the keepers at the end. This is the idea he carries from DEVONthink into AI tools like NotebookLM.
More in Productivity
ProductivityNotion
A connected workspace for docs, projects and AI agents, built around blocks.
16 sources · 74 tips
ProductivityGrammarly
A writing assistant that checks grammar, tone and clarity across apps and browsers.
14 sources · 47 tips
ProductivityDropbox
A cloud storage and file sharing service for individuals and teams.
B12 sources · 28 tips
ProductivityWispr Flow
A voice dictation app that turns speech into polished text across any app.
12 sources · 39 tips
ProductivityOpal
A screen time app that helps people block distracting apps and reduce phone use.
3 sources · 36 tips
ProductivityGamma
An AI tool for making presentations, documents and websites without fixed slides.
4 sources · 39 tips
ProductivitySuperhuman
A fast, keyboard-driven email app for high performers, now with AI features and team collaboration.
10 sources · 29 tips
ProductivityHEY
An email service from 37signals.
B6 sources · 7 tips