the second brain

The second brain - The extended mind.

Attention is the most valuable thing in (human?) cognition, as our high level processing throughput is only 10bit/s (see The Unbearable Slowness of Being)

Note-taking makes thinking incredibly efficient, as you can offload details to persistent memory, allowing your brain to focus on what it does best: Figuring out relationships to abstract concepts.

In combination with a note-taking system, this reduces the task of the brain to identify high level relationships and occasionally look up details at various levels of abstraction.

You offload the bloat to your knowledge management-app, only having to memorize memory addresses and a local cache / working memory.

See also: cached thoughts.

michael levin gets it:

One component to building your own intuition, for having good new ideas, is to show your mind that you’re listening. Again, I have no idea why this works, how generally it works across individuals, or what the mechanism is. All I know is what my experience is and what I’ve suggested to some others for whom it’s worked. The key is that when you get an idea (while jogging, driving, or any other activity), do not let it go and don’t try to just remember it. Write it down somewhere (paper notebook or Evernote software or DevonThink or similar) and go over it later; the key is to get it out of your head, but into a form that preserves it. I’ve noticed that the more you do this, the more the ideas will come. Letting it go and forgetting about it is a signal to your mind that you don’t need ideas; a system that responds to new ideas with a ritual of getting them into a database tends to support their arrival in the future, and frees the creative mind from needing to keep hold of everything in active memory. A close cousin to this is the strategy of offloading all the existing information (ideas, plans, schedules, paper outlines, etc.) into mindmaps, databases, and other tools (see here for more details). I suspect that the more your subconscious believes that the details are safely stored and accessible, the more mind cycles are available for new ideas instead of spending their effort holding on to stuff that’s easily looked up. - src

do not make contemporary physics or other kinds of intellectual endeavour easier, they make it possible. — Neil Levy

Further reading https://www.mentalnodes.com/a-gardening-guide-for-your-mind https://www.mentalnodes.com/from-collector-to-creator https://www.mentalnodes.com/taking-raw-notes-is-useless https://www.mentalnodes.com/folders-kill-creativity https://www.mentalnodes.com/do-not-keep-orphan-notes https://www.mentalnodes.com/the-only-way-to-learn-in-public-is-to-build-in-public

→

https://quartz.jzhao.xyz/philosophy

Think about what purpose the notes serve (in general + on a per-note basis):

Note-taking styles

Textbook: Linear narrative, builds chapter-on-chapter, pedagogical. Reader starts at page 1.

Blog: Standalone posts, polished, public-facing, timestamped. Opinion-heavy.

Wiki: Reference/lookup, neutral tone, comprehensive, heavily cross-linked. Jump in anywhere.

Personal wiki: Wiki but opinionated. “Here’s how I understand X” rather than “X is defined as…“.

Second brain: Memory offload for future self. Distilled essence, not exhaustive. Links carry meaning (not just navigation). Personal perspective baked in.

My vault sits somewhere between personal wiki and second brain. Reference material (definitions, theorems) lives alongside distilled insights and random thoughts. The graph structure matters more than any single note. Notes range from polished explainers to rough stubs that exist mainly to be linked to.

Also a consideration: future AI. An AI that knows what I know can teach hyperpersonalized, collaborate hypereffectively. But also: imagine a “new web” where my AI talks to yours, and to millions of other personal knowledge-bases. Not consensus-seeking but divergent, explorative, open-ended. Like the good parts of twitter on steroids. Each AI distills insights from its unique background and perspective, then reports back: “here’s what I learned today from the knowledge bases I checked out.” AI as information mediator, curator, and collaborator. I think AIs themselves will benefit from maintaining such knowledge bases.

---

My comment on / lessons learnt from Why I Deleted My Second Brain: A Journey Back to Real Thinking

“trying to remember everything” and that’s the core issue, the opposite of how to effectively use obsidian. Following some elaborate note-taking scheme, trying to note eversthing down, …
Your PKM should allow you to do the exact opposite, to let your brain forget, to free it from remembering details, or from the fear of loosing thoughts, connections, references, …
If you try to maintain an encyclopedia, encompassing all information you encounter, you will burn out. But if insteady you take the time to distill the essence of things, from your perspective, for your purpose, … or if you collect references, disambiguations, definitions, etc. — which you would otherwise need to repeatedly search for, just to be presented it in a worse format, provided Google or an LLM can even find exactly what you need without kicking you out of your flow — then your notes help you creatively recombine pointers into thoughts… your notes should help your brain focus on what it does best, by freeing its attention, and offloading details into persistent memory.

Working with static categories, systems, etc. is of course inhibiting. That’s exactly why Obsidian, if used with minimal to no scaffolding around its knowledge structure is so powerful: Just. Flat. Files. Link anything however you want. No rules.

Shareability: I’ve written down and distilled the insight, and can share this compacted idea easily … https://mwolf.dev/general/the-second-brain
Like twitter… sharing memes, but in a more distilled form.

Memory traces: I will update the above and other notes with newer insights, even just attaching this very comment (as is!), not even necessarily for future me to read (in this case!) and rediscover old thoughts / thinking patterns (which sometimes reveals a forgotten perspective on things), but even unreflected dumps of thoughts, interactions, experiences … can not only be useful to yourself, but be processed by LLMs which can index your entire self, or at least a much bigger part of you than without leaving this data legacy, into context. They can reconnect, reanalyze, reprompt, restructure, reformat. More and more effortlessly with time.

There are three main modes of using Obsidian:
1] Reference pasting / collection (low effort, increasingly automatable … in your personal style)
2] Deliberate, reflective, manual, note-taking.
3] Deep introspection… reorganizing notes around a concept, noting down connections (often coincides with 2])

Nested folders, productivity plugins, … all go against these things that make obsidian so useful.

I sympathise with what you say about the “list of unread things”, and creating the illusion of knowledge just because you bookmarked something, and the anxiety of that list growing. “The belief that by naming a goal you are closer to achieving it, that by storing a thought, you have understood it” … also guilty of that. It’s not useful if you take an undisciplined approach.
I also like “I want ideas to resurface because they matter” … but that happens for me with Obsidian too. Just with all the additional benefits listed above. You can more easily filter through and combine ideas.

But the issue of like, “just because it’s written down I now understand it or I can refactor it later or …” is definitely more prevalent when using llms to write notes.
Like I thnk I’ll use claude code less for writing note for the time being.
It’s amazing for papers that I dont actually intend to read in full / detail, wher eI just dont have the time to / they’re not relevant enough (but still interesting / useful, because it benchmarked something or introduces some tangentially interesting idea).
Like then it’s amazing to have a pretty good overview in my obsidian style, have it stored in my vault, have it cross-refernced with other things, that’s amazing.
It’s literally like 20 paper reading. Ok, you read 20% of the paper / spent 20% of the time of reading the full paper, but, like, there’s usually not more inforamtion in most average papers anyways. LIke you get the gist of what they’re tryna say, and the exact methods, math, related work sections are in many cases not worth it / not necessary to get to that core.