Claude Built the Ultimate Second Brain

Aragorn: You already have a cortex. Pippin: what about 2nd cortex?

Job applicants can now start sending their second brain.

I’m building a brain at my work. We have tons of pdf’s from jobs and information collected over 30 years. Every night Claude chews through it all and created corresponding markdown files in a corporate knowledge base. In obsidian too. I teach it also and it creates a markdown to remember all of the tribal knowledge making it captured and searchable.

"We forget most of the things we're supposed to do...." - I can relate :D

The "folders are layers, not topics — topics live in the links" line is the part almost every second-brain setup gets backwards, and it's the whole ballgame for keeping it LLM-legible. But the thing that actually reframed it for me was treating the doctrine as a separate output layer. Everyone builds the wiki and stops at orient — the observe→orient→decide→act framing is what turns a note pile into something that actually acts on your behalf. I've been running a flat markdown + wikilinks memory for a while, and the piece I was missing was exactly that: an immutable raw/ tier feeding a synthesis pass, instead of collapsing sources and conclusions into the same notes. Appreciate you putting the Memex lineage on it too — good reminder this is an 80-year-old idea that finally has a runtime. Thanks Wes.

"Some people have an amazing memory. Hooray for them". GOLD

Doctrine – a term Claude picked up while working at Pentagon 😊

Bro, you make it sound so easy. That's like me showing you tap on guitar solos and telling you 'it's simple, anyone can do it'

I like that this isn't just another "AI productivity" setup. A second brain becomes much more useful when it continuously connects, updates and retrieves knowledge instead of acting like a static notes app. That's where I think these systems really start to shine.

I built a similar thing, but really wanted a "novelty" detector, so I could get something like "in this video Wes is only introducing 20% new content, the rest is stuff he's said before" :) This was helping me screen the far-too-many AI commentaries that are being published on socials. I um-ed and ah-ed about Obsidian but decided on a SQLite database file where I could play with vectors.