by Serhii Kravchenko
For ten years, the technology industry has been building you a Second Brain — apps that store your notes, record your meetings, and now read it all back to you with AI. The idea conquered the world; the products never did. Everything you think can now live outside your head — in systems that hallucinate, lose what you gave them, and break under any serious volume of your thinking. A decade in, the industry is still chasing its own promise. And it will keep missing it, because the failure is not in the engineering — it is in the direction.
In the same ten years, neuroscience and cognitive psychology quietly documented the bill: the better a system remembers for you, the worse you remember. The more it understands for you, the less you understand. We built Artificial Intelligence to carry the mind — and the mind, like any muscle relieved of weight, has been getting weaker under the care.
This document names the alternative. Not a better way to move thinking out of your head — the first serious attempt to aim AI at the opposite target: strengthening the biological brain itself. The one that generates the ideas every Second Brain merely files.
Your First Brain.
Ask yourself where your best ideas arrive. Not where you record them — where they show up.
Nobody says "at my desk." People say: on a walk. In the shower. Running. Driving. Half-asleep. Talking with someone outside at night.
This is measured, not folklore. Across four experiments at Stanford, Oppezzo and Schwartz (2014) found that walking raised the production of original ideas for 81% of participants, while barely affecting convergent, single-answer problem solving. The effect was identical on a treadmill facing a blank wall — it was never the scenery. It is the state itself: the body occupied, the grip loosened, the mind open. Psychology calls the mechanism incubation, and every thinking person on earth knows it by feel.
The brain doing this work is a twenty-watt instrument — less power than the bulb over your desk — and it still produces what no model produces: the original idea, the leap, the judgement, the taste. It remains, by any honest measure, the most remarkable thinking instrument known to exist.
Now look at what we have equipped it with.
Every piece of AI you can buy lives at the desk — a screen, a keyboard, a chair. The state that produces your most original raw material contains no intelligence at all. Which is why the most capable people still walk around with the oldest workaround in the world: a paper notebook. Not out of nostalgia — out of accurate instinct. They know where their real ideas come from, they know no app is present in that moment, and they trust the instrument in their skull over anything built to replace it.
The year's best idea still arrives in motion — and still dies there. And losing those ideas, painful as it is, turns out to be only the symptom. The disease sits in what we built instead.
In 2011, Science published a series of studies by Sparrow, Liu and Wegner measuring what a reliable external memory does to the human one. The result became known as the Google effect, and it is brutally simple:
When people trust that information will be available later, they remember the information less — and remember where to find it more.
The brain is efficient. It refuses to hold what it believes something else is holding. Every act of outsourcing is also an act of quiet deletion.
For a decade, that mechanism has been running against us at industrial scale. Storing, retrieving, and now — with AI reading our material and explaining it back to us — understanding itself has moved out of the head. Each step was a real convenience. Each step also switched off one operation that used to run inside the skull, and every one of those operations was strengthening the brain while it ran.
Cognitive science has spent decades measuring which operations actually build a mind — recalling in your own words, working a thought out yourself, effort placed where it strengthens — and every one of them is an operation these products are designed to remove.
The current generation of AI products is optimised, with great precision, for the exact conditions under which a brain gains nothing.
We bought recall, and we paid with comprehension. That was the trade of the Second Brain decade — never announced, never priced, never put to a vote.
So here is the term, and the alternative it names.
Your First Brain is the one in your skull. For a decade the industry has treated it as legacy hardware — a leaky organ to be backed up, supplemented, and politely replaced. That framing was not just disrespectful. As the science above shows, it was making itself true.
A First-Brain product rejects the premise. The definition is one sentence:
A First-Brain product uses AI to strengthen the mind of the person using it — not to think in its place.
Two directions, and every product ever built for the mind points in one of them:
- Replacement. The AI takes over the thinking — drafts it, decides it, digests it — and hands back the result. The person moves further from their own reasoning with every use.
- Amplification. The AI works on the person's own thinking — catching it at the speed it arrives, in the state where it arrives, carrying it to a finished result, and driving it deeper into the mind it came from. The person moves further into their own reasoning with every use.
Both directions produce impressive products. Only one of them produces stronger people. And this is the standard the term exists to name — a standard the current market cannot quietly adopt, because a product built to think in your place would have to stop doing the very thing it sells.
This idea is not mine, and it is not new. Serious people spent their lives on it.
In 1956, the cyberneticist Ross Ashby named it intelligence amplification — IA, not AI. Not a machine that is intelligent: a machine that makes you more intelligent. J.C.R. Licklider built the case in 1960 with Man-Computer Symbiosis, and in 1962 Douglas Engelbart at Stanford Research Institute turned it into a full research program, Augmenting Human Intellect. Their thesis lost to automation — not on the science, on the economics: a machine that does the work bills this quarter, a machine that strengthens a person pays off over years.
But the deepest reason their road stayed unbuilt is this: amplifying a mind requires a machine that understands thought. Nothing on earth could do that in 1962. Nothing could do it in 2012. LLMs are the first technology in history that can — and they are the opening move, not the destination. A machine that understands what you say is where amplification begins. The full vision reaches further: toward machines that understand what happens as you think. The pieces of that are already leaving the laboratory — non-invasive sensors that read the muscle signals of silent speech, EEG in consumer earbuds, and a map of the brain's own chemistry of motivation and reward, dopamine included, precise enough to build with rather than merely theorise about. None of it reads a mind today, and nothing here claims it does. Taken together, it is the research program of the coming decades — connecting the machine that understands thought to the biology that produces it — and it is not a prediction about someone else's future. It is the direction I have taken up, and this manifesto is its founding document. LLMs are one component of that architecture — the first one powerful enough to start with.
We finally built an intelligence powerful enough to amplify the human mind — and we are spending nearly all of it on replacing the mind instead.
None of this requires implants, mind-reading, or anything from science fiction. It requires pointing AI at moments and mechanisms it has never been aimed at.
At the moment ideas arrive. Intelligence present in motion — on the walk, in the shower, on the run — catching thought at full speed, in your own words, in the state where the mind actually opens. Not a note to process later. The raw material of your thinking, taken in at the only moment it exists.
At the finished result. The distance between a thought and its finished form — the post, the plan, the decision, the script — is where most human thinking dies. Amplification closes that distance to zero: you think out loud, and the work arrives done, in your own voice, with nothing left to process. No backlog, no inbox, no summary to read. Effortless is not a luxury here; effortless is the mechanism, because every unit of friction between a thought and its result is a unit of thinking that never happens.
At the memory itself. This is the frontier, and it is closer than it sounds. Science has twice demonstrated (Rasch et al., 2007; Rudoy et al., 2009) that specific memories can be deliberately strengthened during deep sleep by replaying the right cue at the right moment — a method called targeted memory reactivation. It has been replicated for over fifteen years. It has never once been applied to a person's own ideas. The door between AI and the biological consolidation of your own thinking is already open — and walking through it is the work I have taken up.
This is the arc: begin where ideas are born, close the loop with real results, and press the best of what a person thinks back into the brain that produced it. AI at every step — as the amplifier, never the replacement. And it runs further than software: the same direction, followed for a decade with the seriousness it deserves, leads through neuroscience toward strengthening the First Brain physically — the work of research labs, not feature releases. That is the horizon I am building toward, and the reason this document speaks of an era rather than an app.
The Second Brain was a genuine idea aimed at a real problem. A decade of serious work went into it, and the promise is still unkept — but the problem it saw was real, and nothing here asks anyone to abandon what they use.
But the premise that founded that decade — the mind is for producing ideas, not storing them — was only half honoured. Everything was built for the storing. The producing half, the half its own founders called the mind's real work, was left behind — and for the first time since Ashby named intelligence amplification in 1956, the technology to build it exists. That half is the work of my life, and this manifesto is where it begins.
The term, though, is given away, not defended. If you are building AI that strengthens people rather than replaces them — call it what it is: a First-Brain product. Argue with the definition, sharpen it, hold your work to it. The category grows with every product that earns the name.
Sixty years ago, the founders of computing imagined machines that would make human beings deeper, clearer and more capable — and their century could not build them.
Ours can.
The Second Brain has been built. Now we build the First one.
— Serhii Kravchenko Author of the term "The First Brain" and of this manifesto
References. Ashby, W. R. (1956). An Introduction to Cybernetics. · Licklider, J. C. R. (1960). Man-Computer Symbiosis. IRE Transactions on Human Factors in Electronics. · Engelbart, D. C. (1962). Augmenting Human Intellect: A Conceptual Framework. SRI. · Slamecka, N. J., & Graf, P. (1978). The generation effect. JEP: Human Learning and Memory. · Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning. Psychological Science. · Rasch, B., et al. (2007). Odor cues during slow-wave sleep prompt declarative memory consolidation. Science. · Rudoy, J. D., et al. (2009). Strengthening individual memories by reactivating them during sleep. Science. · Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more learning than elaborative studying. Science. · Sparrow, B., Liu, J., & Wegner, D. M. (2011). Google effects on memory. Science. · Oppezzo, M., & Schwartz, D. L. (2014). Give your ideas some legs. JEP: LMC. · Bjork, R. A. Desirable difficulties in learning and memory (1994– ).
© 2026 Serhii Kravchenko. This manifesto may be shared and quoted freely with attribution.