dBaby

Most second brains are write-only. CORTEX reads them back CORTEX runs on Claude Opus 5.5 inside an Obsidian vault. The clip shows one full loop: capture, distill, link, recall It starts with something messy. A voice memo recorded on a walk: "third breakout this month that faded on me. same thing as the JUP trade. I keep sizing up right after a green week" Opus 5.5 turns it into one clean note. A title, tags, and links to the notes it actually relates to: the JUP trade, sizing after a loss, the rules you keep breaking It links the counterargument too. A note saying breakouts do work in high volatility sits right next to it Then it does what nobody has time to do by hand. It goes through the whole vault looking for connections you never made Two notes about the same rule, written five months apart. A book note and a trade note describing the same mistake. A note that sat alone for 61 days finally gets a home Every link comes with a reason, and every one can be undone Then the question: what have I learned about sizing after a loss? It doesn't match keywords. It walks the graph. Note to note, link to link, 14 notes read in 4 hops The answer comes back with sources. You wrote the same rule twice, five months apart. You broke it twice, both times right after a win The pattern isn't the loss. It's the win before it That's the whole point. Not storing more notes, but hearing what your own notes have been trying to tell you Your notes stay plain markdown in your vault, and every answer points to the notes it came from Independent project built on Claude Opus 5.5 and Obsidian. Not affiliated with Anthropic or Obsidian GitHub: github.com/Shelpid/CORTEX