đ Iâm JA Westenberg. I write essays about technology, philosophy, productivity, and what it means to be human. If you like ideas that take their time, youâre in the right place. You can read my AI disclosure here.
A reminder: if you sign up as a paid annual member, you get a free digital copy of my book Unsanctioned.
If you become a paid Founding member, you also get:
A digital copy of my book Permissionless
One free year of Post Through It
One free year of Kerouac
Upgrade your subscription, claim your perks, and support independent work for the long haul.
Vannevar Bush predicted the modern information crisis in a 1945 Atlantic essay called âAs We May Think.â
He realized scientific knowledge was growing faster than any single brain could handle, and that our organizational systems - rigid alphabetical indexes and library categories - were a terrible match for the way the human mind actually remembers things.
âThe human mind operates by association,â he wrote. âWith one item in its grasp, it snaps instantly to the next that is suggested by the association of thoughts.â
He imagined a device called the memex, a desk with microfilm storage and associative indexing, where a researcher could leave trails through information that would be more valuable than the information itself.
âConsider a future device for individual use, which is a sort of mechanized private file and library. It needs a name, and, to coin one at random, âmemexâ will do. A memex is a device in which an individual stores all his books, records, and communications, and which is mechanized so that it may be consulted with exceeding speed and flexibility. It is an enlarged intimate supplement to his memory.â
The memex was never built, but the idea kept circulating for decades, picking up momentum each time the information environment became denser and more demanding. By the 2010s, when smartphones made access to knowledge ambient and information overload became a clinical complaint, the idea finally found a name that stuck: the second brain.
For most of a decade, people - like me - built them obsessively, maintained them inconsistently, and gradually accumulated a guilty feeling that we were doing something wrong, without quite being sure what.
What I didnât understand then, and think I understand better now, is that the second brain was never really a product problem. It was a layer problem. The first brain has characteristic strengths and characteristic failures. The second brain was designed to compensate for those failures.
But the second brain had its own characteristic failures that no amount of organizational discipline could fully address, and that required a different kind of compensation entirely.
The third brain - AI working on top of both the first and the second - is that compensation.
What the first brain is actually good at
The brain inside your skull is remarkable at things that were required for survival on the African savanna: reading social situations, recognizing patterns from limited data, storing emotionally significant events, and generating creative leaps from incomplete information.
Its working memory holds something like four to seven chunks at once, depending on whose research you find most convincing; its long-term memory encodes through surprise, repetition, and emotional charge rather than through deliberate cataloging. You canât decide to remember something the way you can decide to write it down, which means you remember what moved you, what frightened you, what appeared over and over, and what connected to things you already knew.
Everything else decays.
The good news is that this is what makes human cognition flexible, generative, and occasionally astonishing.
But it also means that serious intellectual work has always needed some form of external scaffold, because it depends on exactly the kind of structured, persistent, arbitrary-sequence-accessible storage that biological memory is poorly suited to provide.
The external and the internal have been entangled for as long as there has been writing. Aristotle described memory using a wax tablet metaphor, the idea that impressions are stamped into the psyche like a signet ring into wax.
"The process of movement involved the act of stamping in, as it were, a sort of impression of the percept, just as persons do who make an impression with a seal [signet ring]. This explains why, in those who are strongly moved owing to passion, or time of life, no mnemonic impression is formed; just as no impression would be formed if the movement of the seal were to strike on running water; while in others, because of the receiving surface being frayed, as happens to [the stucco on] old buildings, or because of the hardness of the receiving surface, the impression does not penetrate."
Andy Clark and David Chalmers formalized this in their 1998 paper âThe Extended Mindâ through a famous thought experiment.
They asked readers to imagine two people navigating to a museum: Inga, who simply recalls the address from memory, and Otto, an Alzheimer's patient who looks up the address in a notebook he carries everywhere. Clark and Chalmers argued that functionally, there is no difference between the two.
When Otto uses his notebook to navigate, the notebook is part of his cognitive system in exactly the same sense that the neurons encoding Ingaâs memories are part of hers. Cognition extends into the world wherever there is a reliable, accessible, and appropriately integrated external resource. The mind leaks outward - and it always has. Writing, numbering systems, and maps are not external aids to thought; theyâre components of thought itself.
The second brain and the decade that wanted it
Tiago Forteâs course, and later his 2022 book âBuilding a Second Brain,â crystallized an anxiety that had been building since smartphones made information access continuous: your biological brain is not equipped to store and retrieve the volume of material a contemporary knowledge worker encounters, so you need a trusted external system that functions as a second brain, capturing ideas, organizing them, and surfacing them when relevant. The framework he called PARA, which organized everything into Projects, Areas, Resources, and Archives, gave people a way to stop improvising their organizational logic and start building something consistent.
The productivity industry latched onto this framing with near-religious intensity. There were arguments about whether to use PARA or Zettelkasten, whether fleeting notes should ever be promoted to permanent notes, whether hand-writing before transcription preserved something essential or just added friction. Roam Research, which launched in 2020, achieved brief cult status by enabling bidirectional linking: if note A links to note B, note B automatically links back, producing graphs of connections that, when visualized, looked something like neurons. Obsidian followed with an open-source, offline-first alternative.
What the second brain got right is that writing changes what you think rather than simply recording what you already think, and that your future self is a different person who will benefit from the archival labor of your present self.
What the second brain mostly failed at
I eventually had to admit: I never actually went back through my notes. I donât think most people do, with any regularity.
The friction of capture is low enough to sustain; the discipline of review, connection, and synthesis was consistently higher than I could maintain. Notes accumulate. Links proliferate past navigability. The graph visualization meant to reveal latent structure starts to resemble a plate of spaghetti. We spend more time organizing our notes than thinking with them, which is the classic Soviet bureaucratic failure: the system designed to support the work becomes the work.
The second brain required you to know, at the time of capture, how your future self might want to retrieve something. You had to tag it, categorize it, connect it to existing notes, name it in a way that your future self would recognize as relevant. But the whole reason you needed the second brain was that your first brain wasnât good at that kind of anticipatory organizational work. You were asking the weak cognitive function, your ability to predict and plan future retrieval, to compensate for another weak cognitive function, your ability to retain and recall past information.
No wonder it felt like running uphill.
The third brain is different in kind
The third brain is AI operating on top of both the first and the second.
AI as a standalone query interface is a good search engine: moderately and impersonally. But AI that has access to your second brain, your notes, your captured thinking, your annotated reading, your past writing, your in-progress ideas, is something qualitatively different. It can do what the second brain promised but couldnât deliver: retrieve relevant material at the moment of need, surface connections you wouldnât have made, synthesize across everything youâve captured, and do all of this without requiring you to have maintained a perfect organizational discipline in the first place.
I have been using something like this for about a year.
When Iâm writing, and I have a half-formed sense that Iâve thought about something related before, I can ask, and it finds it. When I want to know whether my current position is consistent with positions Iâve held in the past, I can ask. When Iâm trying to identify which ideas in my notes are most relevant to something Iâm currently working on, I can ask. The retrieval is imperfect, but it is substantially better than my own memory.
The second brain required you to invest upfront organizational labor in exchange for future retrieval benefit, and most people found the labor-to-benefit ratio unfavorable once the initial enthusiasm wore off. The third brain substantially lowers the upfront hurdle. You can capture messily, knowing that semantic search and pattern recognition will do work that hierarchical tagging used to demand. You can take notes in whatever form makes sense at the time of capture rather than in whatever form will make retrieval possible months later.
Where thinking actually happens now
If the third brain can retrieve, connect, and synthesize, the question of where thinking actually happens becomes murky.
Ottoâs notebook is part of his cognitive system, but it sits passively on the shelf until he consults it, and Otto still does the thinking. The third brain is a processor. It can work on your knowledge base while you are doing something else, making connections you havenât asked it to make, surfacing implications you havenât noticed.
Whatever the third brain is doing while I am asleep, distracted, or working on something else is not something I can consciously identify. And yet the outputs, the connections it surfaces, the syntheses it produces, feed directly into my conscious thinking when I return to it.
Are those thoughts mine?
I have no idea - not yet, at least - and I am suspicious of the confident answers Iâve heard in both directions.
What I have provisionally landed on is that the three layers are less like distinct modules and more like layers of a single extended cognitive Lasagne (forgive me), where each layer compensates for the characteristic failures of the one below.
The first brain is fast, associative, and emotionally intelligent - but painfully forgetful. The second brain is infinitely persistent and organized - but entirely passive, demanding endless manual upkeep. The third brain is a tireless engine for retrieval and synthesis, but it lacks the critical judgment that makes those things valuable. It canât tell you what actually matters or what should change your mind.
But I don't think the goal is frictionless output.
Or at least, I donât think it should be.
We have to intentionally engineer friction into the system, at least on some level, to make sure weâre still actually thinking, instead of reducing ourselves to depressed middle managers servicing an algorithm, living empty lives waiting for the next regional cognitive managerâs yearly conference where we talk about the good old days when we used to have brains and drink s**tty hotel-motel whiskey.