An assistant with your full context.
Reading a paper, you spot a link to something from last week. Ask, and it finds the paper, links your thoughts, writes the note.
Papers, references, highlights, notes, and reading, unified. No more context switching.
Perfect context for AI agents. They work off your tasks, like you would.
Limited beta - Invite only
Reading [[@kaplan2020scaling]] alongside Chinchilla. Two threads worth pulling on.
Compute-optimal frontier shifts when data quality dominates [12]
The "irreducible loss" floor in [7] looks like a measurement artifact
Obsidian for notes. Zotero for references. Acrobat for PDFs. ChatGPT for questions. Plus folders, tabs, bookmarks, screenshots, and citation chains held together by hand. Nothing knows about anything else.
Episteme puts your papers, references, highlights, notes, and reading in one place. One library. One map of what you know.
Read PDFs, write linked notes, and manage references, all in one place.
When your whole research context lives together, you get an assistant that knows all of it, not a chatbot you paste into. It reads, links, highlights, and writes across your entire library. It finds connections you would miss, and does work that was not possible before.
Reading a paper, you spot a link to something from last week. Ask, and it finds the paper, links your thoughts, writes the note.
Add a new paper, ask the agents to highlight every line pink that touches what you already believe. See what is new, what conflicts, instantly.
Need the methodology across fifty papers? Ask once. It works through all of them and hands back a table.
Every claim links back through paper to paper to reference. Trace any idea to its source. Hand your whole library to the next researcher, intact.
What separates a good scientist from a bad scientist, though they are both technically very capable: the great scientist is more creative.
Episteme is in limited beta. Invite only for now.