The post-cognitive era

Before you read any further, let’s take a measurement. 3 questions, you answer to yourself, nobody’s watching.

Question 1: how many phone numbers do you remember? Not counting your own. In the ’90s, your parents carried a dozen to several dozen in their heads — because they had to.

Question 2: when did you last drive an unfamiliar route without navigation — from a map you held in your head?

Question 3: when did you last calculate anything more than a tip without reaching for your phone?

If these questions sting a little — good. That wasn’t a memory test or a nostalgia quiz about “the good old days”. It was a measurement of the boundary of your mind.

Chapter 1 left us with Clark and Chalmers’s thesis: the mind doesn’t stop at the skull, and Otto’s notebook was part of his cognitive system. It’s easy to nod along and file it on the shelf of philosophical curiosities. It’s harder to notice that you just tested the thesis on yourself. Let’s call it the subtraction test: to find out what truly belongs to your cognitive system, subtract it — and measure the difference. Subtract Otto’s notebook: he stops knowing where the museum is. Subtract your phone: you stop knowing numbers, routes, and dates. Functionally, it’s the same amputation.

And the most interesting part is what you don’t feel in this measurement. Day to day, you don’t experience navigation as “using a tool” — you experience it as finding your way. That is exactly the automatic-trust condition from chapter 1: the tool has grown in so deep that it has disappeared.

Philosophy named this state a hundred years before the first smartphone. In Being and Time (Sein und Zeit, 1927), Martin Heidegger observed that a tool exists for us in two ways — and the German language gave him two words for it. A hammer in motion is zuhanden, ready-to-hand: it vanishes from awareness, it’s an extension of the arm — you see the nail, not the hammer. It becomes vorhanden, present-at-hand, only when the handle snaps: suddenly it’s an object, a problem, a thing to examine. We talk about a tool only when it breaks — as long as it works, we simply live through it. Your navigation is zuhanden. And that’s why the subtraction test works at all: it is nothing but a controlled snapping of the handle — the only known way to see the hammers you think with.

This chapter turns that observation into a definition — an operational one, meaning one you can measure rather than merely recite. Along the way: the full version of the 4 waves of extension, the world’s oldest technology panic (it’s 2,400 years old and still current), and the 8 traits by which you’ll recognize the era — in yourself, in your company, in society.

The operational definition

Every word of this definition is doing work, so let’s take it apart layer by layer. Human — biology remains the foundation and the integrating layer. Thinking model — the concepts and frames you cut reality with; the very notion of a “protocol” from chapter 1 is such a layer: once you have it, you see something different in a shopping mall than you did before reading. AI — the layer that performs operations on meaning. Data — what your tools know about your world. External memory — from Otto’s notebook to the memory of your conversations with a model. Composite cognition is the output of this entire stack, not of biology alone.

What makes this definition operational and not poetic? It has criteria. 3 thresholds — and notice that they are exactly Clark and Chalmers’s conditions from chapter 1, turned into control questions:

The subtraction threshold (constant availability): subtracting the non-biological layers degrades your cognitive performance — in a growing number of tasks, more than subtracting a teammate would.

The trust threshold (automatic trust): you don’t routinely verify the stack’s answers, just as you don’t routinely verify your own memory. (Whether that’s wise — a separate matter; we’re measuring whether it is.)

The invisibility threshold (instant access): you stop noticing the moment you reach for the tool. Nobody says “I will now make use of the satellite navigation system”. People say “hold on, let me check” — the tool fully zuhanden.

An individual crosses these thresholds privately. A society crosses them when institutions start to assume the stack: the school assumes the student has a search engine, the employer assumes the employee has a model, the government office assumes the citizen has a phone. At that point the composite stops being a choice — it becomes the default configuration of a human being.

An attentive reader will object: this definition also catches a calculator from 1975. Correct — and that’s the point. The era doesn’t start from zero; it starts when the stack is staffed in full. The calculator staffed a narrow slice of one layer. The search engine staffed access. Only the LLM staffed the last and most general layer — processing itself — and across all symbolic domains at once. To see why this is a threshold and not just another step, we need the full version of the table from chapter 1.

The four waves — the full version

Chapter 1 showed the 4 waves of extension in compact form. Now let’s add the 2 columns that do all the work: who loses a monopoly, and how fast the wave spreads.

Two things become visible only in this version of the table.

First: every wave takes away somebody’s monopoly on mediating knowledge. Writing took it from the living archives, print — from the copyists and the censors of copying, the internet — from the gatekeepers of access. The fourth wave does the same to the last monopoly left: the expert as the only place where reasoning happens.

Second: the waves aren’t parallel — they’re a stack. Print needed literacy, the internet needed print and telecommunications, AI needed the internet twice: as the source of training data and as the distribution channel. Each wave rides on the infrastructure of the previous one — and that’s why adoption accelerates by an order of magnitude every time.

The oldest panic in the world

With every wave, the same cry goes up: “this will make us stupid”. It’s worth knowing that the cry is 2,400 years old and has an excellent pedigree.

In the Phaedrus, Plato puts into the mouth of the Egyptian king Thamus a reply to the god Theuth, the inventor of writing: this invention will implant forgetfulness in souls, because people will stop exercising memory — they will recall from external marks, not from within themselves; they will possess the appearance of wisdom instead of wisdom. Sound familiar? It is exactly today’s charge against LLMs, word for word, only about writing. (The Socratic irony: we know this argument solely because Plato wrote it down.)

The most important thing about this story is not that the panic was silly. It’s exactly the opposite: the panic was half right. The art of memory really did die — oral cultures, epic reciters, memory palaces are folklore and hobby today. Print really did flood Europe with errors and propaganda — the religious wars of the 16th-17th centuries were fought on printed pamphlets. The internet really did change memory: the Sparrow, Liu, and Wegner study (Science, 2011) showed that when we know information will be available, we remember where to find it, not what it says. Nicholas Carr asked in 2008, “Is Google Making Us Stupid?” — and it was not a stupid question.

Every wave genuinely amputates something. The question of the era is never “are we losing something” — we always are. The question is: is the trade worth it, and who controls its terms. To the first part of that question, history has answered “yes” three times. The second part — who controls the terms of the trade — is the subject of chapter 9, and it is the first part for which the answer is not a foregone conclusion.

What the fourth wave does differently

The first three waves, for all their power, shared a ceiling: the tools stored and transported records of thinking, but every operation on meaning — comparison, inference, synthesis, critique — happened inside a skull. A library could hold all the books in the world; a human had to read and connect them.

Wave 4 breaks through that ceiling with three properties at once.

It performs operations on meaning. The model doesn’t hand you a record of someone else’s thought to process on your own — it reasons, summarizes, critiques, translates between levels of abstraction. For the first time in history, part of the processing happens outside biology.

It answers back. A book won’t ask you a follow-up question, a search engine won’t say “notice that your assumption contradicts what you wrote earlier”. Conversation is a different category of interaction than reading — and the first 3 waves didn’t have it.

It is general. The microscope extended seeing, the calculator — arithmetic, navigation — orientation. One tool, one function. The LLM is the first general-purpose cognitive tool: the same layer handles law, code, marketing, and a letter to a government office.

At this point someone will rightly raise a hand: what if all of this is just a statistical parrot that doesn’t understand anything? Maybe. For an operational definition, though, it makes no difference — and that is a convenient property of operational definitions. The subtraction test measures the output of the human+model system, not the metaphysics of the model. If subtracting a layer degrades the output, the layer is real — regardless of whether it “really thinks”. We leave the dispute about true understanding to philosophers of mind; for us it’s enough that the difference is measurable. The microscope from chapter 1 didn’t “understand” bacteria either — and microbiology happened anyway.

The 8 traits of the era — how you’ll recognize it

The definition and the waves say what happened. The 8 traits say how you’ll know it — treat them as an observation list, not a catechism. The numbers in this section are stylized orders of magnitude, not laboratory measurements; the observables can be checked with the naked eye.

1. Asymmetric cognition. Two people with identical education and tenure now differ in output not by percentages but by orders of magnitude — depending on whether one of them has built a composite stack. No previous educational technology stratified this fast: a diploma differentiated over decades, a stack differentiates within quarters. Observable: “one-person departments” start appearing in companies — and nobody knows how to slot them into the pay grid.

2. Throughput expansion. One person with a stack does work that recently required a team and months. Observable: projects that “weren’t worth it” at the old costs of execution suddenly are — and someone is doing them solo.

3. The second death of memorization. The first death (the internet) was about facts: after Sparrow and Wegner, we know that we remember where, not what. The second (the LLM) is about procedures and patterns: how that kind of contract gets written, how that kind of analysis gets done, what the code of that kind of module looks like. What stays in the head is a map, not a warehouse — knowing that something exists, when to use it, and how to recognize a good result. Observable: experts increasingly say “I don’t remember the syntax — I remember what to demand”.

4. The disintermediation of authority. Authority shifts from credentials (years of tenure, title, diploma) to results. A junior with a stack regularly delivers things that yesterday required a senior — which doesn’t mean experience has lost its value; it means it has stopped being the indispensable intermediary between problem and result. Observable: the portfolio beats the CV; the client says “show me”, not “how many years”.

5. Coevolutionary symbiosis. You train your stack — with memory, context, corrections. The stack trains you — the patterns it serves you shape your own. After a year of this loop, your composite is unrepeatable: a mental fingerprint. The same loop has a shadow — if millions of people train on identical default settings, their fingerprints grow alike (chapter 9 calls this the homogenization of thought). Observable: your stack performs distinctly worse in someone else’s hands — and vice versa.

6. The compression of expertise — and its hard limit. In chapter 1 we promised to settle what exactly gets compressed. What compresses is codified knowledge: syntax, patterns, boilerplate, the textbook state of a field, translation between formats and levels of abstraction — everything that was ever properly written down, because that is precisely what the model learned from. What doesn’t compress is what was never properly written down: judgment about what matters; taste; responsibility for the outcome — the signature a model will not give; the local context of your company and your people; trust; the ability to ask a question nobody has asked yet. The famous 10,000 hours don’t disappear — they change address: the hours spent on “how” move into the tool, the hours spent on “what and what for” stay with you and get more expensive. This is the demand inversion of the era: the market stops paying a premium for I know how — the stack can do that — and starts paying a premium for I know what not to do. Observable: the value of people who close options, rather than open them, is rising.

7. Distributed authorship. Who wrote this code, this analysis, this chapter? A human + a model + data + editing. The classical notion of authorship — one head, one pen, one signature — is cracking, and law, school, and science are improvising patches. Observable: multiplying AI co-authorship disclosures. The footer of this book is one of them — not out of obligation, out of craft.

8. Hyperliquidity of intent. In finance, liquidity tells you how fast an asset converts into cash. Hyperliquidity of intent tells you how fast an intention converts into a working artifact: idea → prototype in hours, not months. Since execution is getting cheaper by orders of magnitude, it stops being the bottleneck — and the bottleneck becomes selection, because attention, unlike execution, hasn’t gotten one bit cheaper. The scarce resource of the era is not “knowing how to build”. It is choosing what not to build. Observable: backlogs grow faster than the capacity to think them through; “anything can be built” turns from a boast into a management problem.

Three lines, one conclusion

Finally — the pedigree. In chapter 1 we promised that “post-cognitive era” doesn’t hang on a single name. It hangs on three independent philosophical traditions that for decades barely read each other — and arrived at the same conclusion from three different directions.

Line one, philosophy of mind: Clark and Chalmers, familiar from chapter 1. The mind doesn’t stop at the skull; in 2025, Clark personally carried this line all the way to generative AI.

Line two, philosophy of information: Luciano Floridi (The Fourth Revolution, 2014). Floridi counts the revolutions that stripped humans of their central place: Copernicus — we are not the center of the cosmos; Darwin — we are not separate from the rest of life; Freud — we are not masters even of our own psyche; and finally Turing — we are not the only entities that process information. Hence his “hyperhistory”: prehistory had no records, history had records, hyperhistory is societies whose functioning depends on information technologies — and “onlife”: a life in which the boundary between online and offline stops being perceptible.

Line three, philosophy of technology: Bernard Stiegler (Technics and Time, from 1994). Stiegler distinguishes three retentions: primary (what perception holds in a given moment), secondary (what you recall from your own past), and tertiary — memory exteriorized in technics: writing, photography, tape, the database, the model. His thesis is stronger than Clark’s: technics is not an add-on to a ready-made human — the human is constituted, from the very beginning, through technical prostheses. In this language, the LLM is a tertiary retention of a new generation: the first that doesn’t just store past thinking but actively replays it.

Technics is the pursuit of life by means other than life.

Technics: the continuation of life by other means. Anglophone philosophy of mind, Italian philosophy of information, French philosophy of technology: three schools, three methods, three starting points — one conclusion. The boundary between the mind and its technical environment is conventional, and it is moving right now. When three independent survey lines intersect at the same point, the cartographer stops asking whether the point exists and starts charting it. This book is that charting.

The era doesn’t ask for consent

Let’s pull it together. The post-cognitive era has an operational definition (a composite stack + 3 thresholds: subtraction, trust, invisibility), a pedigree (4 waves, of which the fourth is the first to externalize processing), and a list of observables (8 traits). You have the tools to check for yourself whether the thesis holds up against reality — considerably more than the average buzzword offers.

Notice, too, what this definition does not say. It doesn’t say you should be pleased. It doesn’t say the composite beats the lone head at every task. It doesn’t even say this ends well — that’s what chapter 9 is for. It says only: this is already happening, it can be measured, and you cross the thresholds whether or not you gave consent. Your children’s school, your employer, and your government office already assume the stack.

If cognition has become composite, the advantage shifts to the people who deliberately design their composite — and who see the protocols their composite flows along. In chapter 1 we called them protocol-era operators. Chapter 3 takes that configuration apart: 5 skill stacks, historical precedents, and a test for whether — just maybe — that’s you.

The subtraction test works in both directions. You can measure what has already grown into your mind without your consent — or start consciously choosing what you let in. The first is a diagnosis. The second is a design. This book is about the second.

---

The post-cognitive era — the period in which cognition stops being an exclusively individual resource and becomes composite: human + thinking model + AI + data + external memory. An extension of the Extended Mind thesis (Clark & Chalmers, 1998) into the age of LLMs.

Methodological disclosure: this book is written with AI as a co-author — this chapter was written by Claude Fable 5 (June 2026) from the author’s conceptual framework, with facts and quotations verified at the source; this English edition was translated from the Polish original (June 2026). This is not a gimmick but consistency with the thesis: a text about composite cognition is written by composite cognition — and thinking is versioned the way code is.