stash/docs/ai-native-bookmarks.md at main · Fergana-Labs/stash · GitHub
//blob/show" data-turbo-transient="true" />
Skip to content
Type / to search
Sign in<br>Sign upAppearance settings
You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.
Dismiss alert
{{ message }}
Uh oh!
There was an error while loading. Please reload this page.
Fergana-Labs
stash
Public
Notifications<br>You must be signed in to change notification settings
Fork<br>50
Star<br>294
FilesExpand file tree
main
/ai-native-bookmarks.md
Copy path
Blame<br>More file actions
Blame<br>More file actions
Latest commit
History<br>History<br>History
132 lines (96 loc) · 7.5 KB
main
/ai-native-bookmarks.md
Copy path
Top
File metadata and controls<br>Preview
Code
Blame
132 lines (96 loc) · 7.5 KB
Raw<br>Copy raw file<br>Download raw file
OutlineEdit and raw actions
AI-Native Bookmarks: the private internet library
How it works
A bookmark manager whose reader is an agent, not you. Give your agents a personal<br>internet filled with all the best things you've opened, or never got around to, on the<br>inter-webs. Three steps.
1. Save anything from the web
Add to Chrome — it's free
A Chrome extension that saves websites, PDFs, YouTube videos, X posts, Instagram reels,<br>and ChatGPT or Claude chats straight into Stash. Automatic. Searchable. Read by your<br>agents.
Clip any page — articles, PDFs, and every open tab. Saved clean and readable.
Bring your bookmarks — import the whole file. We fetch what's behind every link.
Twitter bookmarks — your X bookmarks, synced. Text, images, and threads kept.
AI chats — ChatGPT and Claude, streamed in. Searchable like everything else.
Saving captures the content , not the URL, so a dead link still has the article behind it.
2. We maintain a private library and build an LLM wiki of key topics
We save everything to a private library that's filterable and sortable. The<br>saved copy stays readable even when the original goes away. Stash reads what comes in and<br>keeps a wiki of the topics you keep returning to, with each entry still attached to the<br>page it came from.
3. Your existing agents use Stash to get context right away
Claude, ChatGPT, and Cursor reach the whole library through the CLI or MCP, so they answer<br>out of what you've actually read. No new chatbot to switch to.
What early users are doing with it
Keeping on top of the deluge of AI papers.
Making sure articles and books actually change something, instead of being consumed and forgotten.
Studying for exams; collecting sources for essays.
Maintaining situational awareness of an ecosystem or market.
The case for a personal library
There's too much to consume as a single person. Between social media like tiktok and<br>twitter, news articles, youtube channels, and podcasts, we are suffocated with<br>information. The process of choosing what to read is overwhelming, never mind doing the<br>reading itself.
Google's original mission was to organize all of the world's information. Along with the<br>other companies that pioneered the internet, this became a race to gobble up as many data<br>sources as possible in order feed the machine. Along the way, the stacks of the library<br>became too numerous.
LLMs, with a little help from Jevon's paradox, is going to make this problem worse. In a<br>week since starting, our summer intern has generated 10M tokens or 100 books. Multiply<br>that across everyone on the internet and you have an enormous problem for anyone trying to<br>keep on top things. How do you know what to spend your preciously limited time actually<br>reading? The fact that LLMs tend to be quite verbose makes this worse as the density of<br>information across writing decreases and AI slop proliferates.
What to do
Choosing to read only from trusted sources is one solution to this. Thoughtful, articulate<br>voices with reputations to stake will become ever more important over time. But this<br>cannot be a complete solution. Such a world would undermine the marketplace of ideas that<br>powers free speech and democratic societies. Smaller voices would be shut out and society<br>would become less diverse, less informed. It would lead to a form of mode collapse,<br>brought about by human vices rather than technical failure.
The solution, we believe, is to enable agents to be the intermediary. Agents read<br>everything and filter down to only what is worth reading. They can serve up all the<br>relevant information at the right time so that decision makers are fully informed. They<br>can serve as thought partners that are not necessarily smarter, but more informed than any<br>individual human can ever be. Whether that's making sure all the latest papers can be<br>brought to bear on a new research direction, resurfacing impactful perspectives in a<br>moment of personal crisis, or flagging hidden concerns during a major business decision.
In a lot...