Ask HN: How would you build second brain in the AI era?
A lot of people use 'second brain' software and are now trying to connect/integrate them with LLMs.
But if you were starting with a new 'second brain' from scratch, how would you do it, given the current state (and trends) in AI/LLMs?
I think it depends on what the person does. I'm a knowledge worker, so this works for me. Other people may see little use in notes.
A pool of well-structured notes and knowledge with an "AI technologies du jour" trained on it.
I currently use Obsidian and some tools for this, although it's not completely frictionless yet.
Less structure? I think links are less necessary. It's a bit like using email when search is cheap, you don't need to rely on tags if you don't want to.
I have a fairly large set of notes written in Org, and I never bothered with internal links. Ripgrep search was always enough for me to make connections.
With local LLMs, it's even better. gptel makes it really easy to add notes as context.
I wouldn't. It's a vulnerability to have your data somewhere else.
People know this. The first actual "second brain" of some sorts is likely to be unwilling, made by someone else trying to capture the thoughts of a subject. It is also likely that it will fail (given the fact we don't know shit about how brains work).