AI and Me: Building My Second Brain
This week has been all about my new ‘second brain’.
This is a bit like the catalogue cards in a 1970s library being morphed directly onto your own screen. I am a lover of knowledge: different blocks of information, processed by biological intelligence and stored to connect with other bits of information, drawing insights and setting the light bulbs flashing. The biological process is highly dependent on bandwidth and ability to retrieve. How tired I am, how much is going on, or the risk of losing a thread if someone has finally had enough of me.
In recent times, I have been learning from references to Karpathy’s work (a renowned researcher and teacher in AI). So I have built myself a ‘second brain’. Easy with Claude Opus 4.8.
I identified my key data sets: infrastructure, covering extensive work involving power, underground water services, transport, development contributions, and so on; and project information, where much of it is cross-relevant.
It is set up by major theme under my Claude folder: 2nd Brain / Knowledge Base_Infrastructure. Each knowledge base has its own claude.md file and a folder each for Outputs, Raw, and Wiki.
It is running to self-improve and update itself with any new additions at 2am every day (because it has to load). In one of my databases, under one infrastructure provider, I have saved about 1,200 items.
I will be interested to see how this runs for me. I suspect I now have my own private librarian and a powerful knowledge system that is mine.
Note: I found a large amount of information from 2024 and 2025 that was buried. It was public information but not readily available. In fact, one item I requested had the Council coming to me to find out where I had got it so they could retrieve the appendices.
If you are new to Cowork, how could you use this type of knowledge management?
- Customer and client database sets
- Projects, so everything is together and you can search readily across them
- Managing consent processes
- Managing and recalling contract instructions (simple and small, but a good use)
Once this is running, I will ask Claude to prepare a SoP covering how I built it and how I will be using it. I will share this with my close peers who I am on the AI journey with.
I remain the sovereign of my own mind. I seek enhancement in my outputs through carefully considered use of AI.
Thanks for sharing, Andrej Karpathy, and to the others who are sharing too. It really helps.