The thinking & ideas system
This is my favourite part of the system, and the reason the rest of it exists. The problem it solves: I think constantly, walking, reading, at 2am, and almost none of that used to survive. Now all of it does, and it comes back to me connected. I can link many more ideas than before, and the system has become a thought-partner: it spots linkages I don't, and it frames and structures ideas in ways that let me spot linkages it doesn't. This system is the sole reason for many a 5am bedtime over the last few months.
A thought becomes a finished note without me doing anything else. I speak into my phone, into an ideas channel or a journal channel, or I type. By the end of the day that capture has become a properly written note: journal fragments turned into full sentences in my own voice with nothing dropped, each idea filed as its own note, everything dated, tagged, and cross-linked into the rest of the vault. The raw recording is kept verbatim forever, so the polished version never has to be trusted blindly.
My journal works for me while I sleep. Journal entries do not just sit there; they get mined. A task I mentioned in passing lands in my task system. Something I said about what I want, or where I am heading, feeds my goals layer. A book someone recommended or a restaurant I loved gets extracted into the taste system. A stray idea gets split out into the ideas lane. Journalling stopped being a diary and became an input into everything else I run.
Everything I have ever written is one searchable library. Essays, ideas, journal entries, working documents: one library, tagged by type and by my standing intellectual interests, among them AI, consciousness, and rationality. When I want everything I have ever thought about consciousness, I get the essay I published, the half-formed idea from a walk, and the lecture where it came up, together, even though they arrived through completely different doors.
Everything I read and listen to lives here too. The library is not just my own writing. Every article, podcast, and video worth keeping goes in through one capture shortcut on my phone, and arrives fetched, summarised, tagged, and filed, linked to its source and to everything related already in the vault.
An example flow. It runs end to end without me touching a file: I finish a podcast on a walk, share it to the capture shortcut, then speak an idea note with my reflections on it. By the evening the podcast is filed and summarised, my reflections are written up as an idea note linked to it, and every concept either one touches is linked to the related essays, articles, and thoughts already in the library. Weeks later, when I am thinking about the same topic, I ask Claude one question and get back a synthesis of every related concept, framework, and idea I have ever captured: the reading, my reactions to it, and the connections between them, assembled in seconds.
I can interrogate what I have studied. I have a set of consultants: AI agents, each grounded in a specific body of work I have actually engaged with. One is built on the LessWrong Sequences and stress-tests my reasoning when I have a decision to make. Others are built on my MBA coursework, so I can ask my negotiation course a question the night before a real negotiation. Another is grounded in the published writing of a thinker on identity and leadership whose work I studied. When a question is big enough, I convene several of them at once and have them argue: each critiques the others' positions blind, and I get a synthesis that keeps the disagreement visible instead of averaging it away.
Everything connects itself. Every note carries a live block showing what it links to and what it overlaps with: shared people, shared topics, shared threads, ranked. I do not maintain any of it.
The loop underneath is capture, distil, connect, interrogate: catch everything cheaply, turn it into writing worth keeping, let the connections assemble themselves, then question the accumulated whole as if it were a colleague who has read everything I have ever read and written, and forgotten none of it. The pipelines, schemas, and scheduled jobs that make this run are documented in the infrastructure pages; this page is what they are for.