31 Glossary
(ns meme-constructor.posts.glossary
(:require [scicloj.kindly.v4.kind :as kind]
[meme-constructor.glossary :as g]
[meme-constructor.lexicon :as lex]))The terms of the book, in alphabetical order. Each entry says what the term means here, where it comes from, and where the book introduces it. The links under “Introduced in” lead to that place. The lexicon lists the short codes; where a term belongs to a coded item, the entry names it.
In the posts, a term in italics links to its entry here. The entries are generated from resources/glossary.edn. The “also used in” lines are computed when this page is rendered, by scanning the prose of the introduction and of every post of Parts I to III for links to the entry.
31.1 A
31.1.1 Abstract constructor
Information that meets the definition of a constructor better than any of the physical things that carry it. Deutsch’s example is the Ship of Theseus: the planks, the plans and the Athenians all change, and what persists and keeps the ship in repair is the information in the plans and in the ideas (Deutsch 2013, §2.14). Marletto (2015) uses the term for genetic information. Post I.1 makes it precise for programs: the class of all programs that perform the same tasks, whatever their text. A meme is an abstract constructor in this sense.
Introduced in: introduction, post 1, post I.1
Lexicon: meme as equivalence class (F2)
Also used in: post I.0
31.1.2 Accuracy
How close a device comes to a task in a single run: the distance between the output attribute the task asks for and the output the device actually produces, “when run once” (Deutsch & Marletto 2025, §2, following Violaris & Marletto 2022). Which distance to use is up to the particular theory. Accuracy is one of the two ways a device can fall short of a constructor. The other is reliability.
Introduced in: introduction, post 1
31.1.3 Adaptationism
The conjecture that adaptationism is the best theory of design (post I.5, Conjecture I.5.12): every constructor that reliably produces a design without containing it implements unspecific variation and selection by the task. Competition is not needed. It is supported by showing that its rivals, such as spandrels, developmental constraint, self-organisation and exaptation, say nothing positive about design: they fail the alternatives condition or explain no fine-tuning. In Part III, explanatory adaptationism is its biological instance (post III.1, after Godfrey-Smith 2001). It says nothing about how many traits are designs, and it is unproved.
Introduced in: post I.5, post III.1
Lexicon: explanatory adaptationism and its critics (AD2)
31.1.4 Adaptive process, adaptive synthesis task
A synthesis task carried out by a population: a task, a medium of candidate programs, a variation constructor, an elimination constructor and a finite resource (post I.5, Definition I.5.1). It has two conditions: the variation must not read the task, and which variants persist or get varied must depend on how well they perform it. Competition for the resource is a property some adaptive processes have, not a condition of design. Evolution by natural selection and the growth of knowledge are both adaptive processes.
Introduced in: post I.5
Lexicon: adaptive synthesis task (C5), unspecific-variation condition (A1′), competition (A2), selection condition (A3)
31.1.5 Anchor
What a ranking used in keeping variants depends on (post 4, Definition 4.2). The ranking may be the criterion’s, deciding which candidate a host keeps, or transmission’s, deciding which host gets copied. The anchor is internal if the ranking depends on the meme or on its hosts, as with closeness to the received version or the hosts’ psychology. It is external if the ranking is the same for every meme and every set of hosts, as a test against reality is. Under an internal anchor the content is blind to reality. The book proposes that rational and anti-rational memes differ in their anchor.
Introduced in: introduction, post 4
Lexicon: criterion and anchor (F8)
Also used in: post 0
31.1.6 Anti-rational meme
A meme that, in Deutsch’s words, survives “by disabling [its] holders’ critical faculties” (Deutsch 2011, ch. 15). This book argues that anti-rational memes also use the creativity of their hosts, since copying a culture meme needs it. They differ from rational memes in the anchor of their criterion, which is internal: the meme itself or its hosts (post 4). A good anti-rational meme is a good mind virus.
Introduced in: introduction
31.1.7 Ape meme
A meme whose repertoire is fixed by a string on the closed level. Copying has two levels. The closed level is the channel (sounds, words, writing, a fixed set of behaviours): a closed alphabet that is fixed in the receiver at the time of copying. The fixed alphabet is the central property of this level. The receiver parses the observed behaviour into the alphabet, so deviations smaller than the distance between two symbols are absorbed, and the copy can be made arbitrarily accurate (the receiver’s behaviour parsing is Byrne’s; the arbitrary accuracy is our claim). The book conjectures that the symbols of the alphabet are attractors in Sperber’s (1996) sense, and that they explain why copying on this level is faithful. Its limits are that it is expensive to pass on and expresses only what the alphabet composes (Deutsch 2011, ch. 16). Recitation, checklists and forks of source code are the human cases (post 1, Section 3.1). Ape memes and culture memes are not a partition of memes: the same channel can carry a culture meme, whose function lives on the open level.
Introduced in: post 1
Also used in: introduction, post 0
31.1.8 Approximate constructor
A real device that performs a task with less than perfect accuracy and reliability (Deutsch & Marletto 2025, §1–2). Catalysts and heat engines are examples. A task is possible if there are approximate constructors for it that come arbitrarily close to perfection. The devices themselves need not converge to anything.
Introduced in: introduction, post 1
31.1.9 Aspect
What a criterion writes about a hypothesis, such as “fails test 3” or “contradicts the evidence” (post I.6, Definition I.6.14). An aspect may lead to a rejection, or may only be information about the hypothesis: a new problem.
Introduced in: post I.6
Lexicon: critics (C11)
31.1.10 Attribute
A set of states of a substrate, such as “red” or “contains the sequence ACGT” (Deutsch 2013). A meme is an attribute of brains: the set of brain states that carry it.
Introduced in: introduction, post 1
31.2 B
31.2.1 Behaviour parsing
Copying by observation without understanding: the observer picks out recurring patterns in the stream of behaviour and copies its hierarchical organisation (Byrne 1999, “string parsing”; Byrne and Russon 1998, “program-level imitation”). Mountain gorillas prepare nettles in a standard sequence of steps, although the hand movements within the steps differ between animals (Byrne and Byrne 1993). We read the parsing as running into a fixed alphabet of elements the receiver already has, which is what makes the copy of an ape meme digital and arbitrarily accurate. That reading is ours, not Byrne’s.
Introduced in: post 1
31.2.2 Bucket theory, searchlight theory
Popper’s two pictures of the mind (Popper 1972, appendix 1). On the bucket theory the mind is a container into which experience pours. On the searchlight theory the mind has expectations first, and they decide what it looks at. Popper rejected the bucket. If the bucket theory is false, an idea cannot be poured from one head into another: the receiver has to rebuild it, and what it is shown can only criticise its guesses.
Introduced in: introduction
31.3 C
31.3.1 Circuit
A physical system of gates, each performing an elementary task on a few bits, wired so that the laws of physics carry out a computation: logic gates in a classical computer, a quantum circuit in a quantum computer (post I.1, §1.2). It is the bottom of every software stack, and nothing in it is interpreted.
Introduced in: post I.1
31.3.2 Computing system
A substrate some of whose variables are information variables, among them an input medium and an output medium, and whose patterns act on the system itself (post I.1, Definition I.1.1). The definition keeps the system’s own steps apart from the steps caused from outside, such as inputs and programming.
Introduced in: post I.1
Lexicon: computing system (C1)
31.3.3 Conjecture and refutation
Popper’s account of how knowledge grows: we guess, and then we try to find out what is wrong with the guess (Popper 1963). The book reads the copying of a culture meme as conjecture and refutation in the receiver.
Introduced in: introduction
31.3.4 Constructor
Something that performs a task and keeps the ability to perform it again (Deutsch 2013). A catalyst, a factory and a computer running a program are examples. Strictly, a constructor is an idealisation: a device that performs its task with perfect accuracy and perfect reliability. Deutsch & Marletto (2025, §2) call it a “fictitious ideal device”. Every real device is an approximate constructor. The book uses the word only in this technical sense.
Introduced in: introduction, post 1
Lexicon: constructor-theory basics (F1)
31.3.5 Copier
The part C of a vehicle that copies the recipe (Marletto 2015). A copier is generic if it copies every content in its range equally well, as DNA replication does, and content-specific otherwise (post 1, Definition 1.15). The book conjectures that meme copiers are content-specific (Conjecture 1.16).
Introduced in: post 1
31.3.6 Copy task
The task that takes evidence for an original, finitely many of its answers, and a receiver to the same evidence and a receiver that performs every task of the original (post I.3, Definition I.3.7). The copier sees only part of what it has to rebuild, so the copy task is synthesis from examples with a weaker specification than the one it has to meet. For a meme \(m\) it is the task \(C_m\) of post 1 (Definition 1.8), whose error splits into a loss ε⁻ and a gain ε⁺.
Introduced in: post I.3
Lexicon: copy task and ε (F3)
Also used in: post 1
31.3.7 Creativity
The receiver’s own process of conjecture and criticism. For an open receiver no finite evidence fixes the idea behind it, so the constructor that performs the copy task of a culture meme must propose conjectures of its own: it is the receiver’s creativity (post 1). Ape memes need none, since the receiver parses them into a fixed alphabet. Deutsch (2011, ch. 16) argues that creativity was first selected to copy memes, not to invent them.
Introduced in: post 1
Also used in: introduction
31.3.8 Criterion
The map that decides, when a meme is copied, which variants are kept and which are discarded (post 4, Definition 4.1). Post I.6 splits this into two parts: a criterion in the narrow sense labels a variant with an aspect, and a critic adds the decision (Definition I.6.14). The book considers criteria that rank each candidate against the meme it came from and keep it if it ranks at least as high.
Introduced in: post 4
Lexicon: criterion and anchor (F8)
Also used in: post I.6
31.3.9 Critic
A task from a hypothesis to a verdict, reject or keep or a rank, that has a criterion as its subtask: first the criterion labels the hypothesis with an aspect, then a decision reads the aspect (post I.6, Definition I.6.14). A critic acts on a representation of the hypothesis (Conjecture I.6.15). It can refute, never certify (Proposition I.6.16).
Introduced in: post I.6
Lexicon: critics (C11)
31.3.10 Culture meme
A meme that is carried by a closed channel (sounds, words, writing) but whose function lives on the open level. The function is not a string. Two heads share the signal alphabet but not the concepts, so the receiver rebuilds the function, in the open mode, from the evidence and from reasons with its own creativity. An utterance or text is a recipe-like signal that the receiver reads as evidence and does not execute. No exact copy of the function can be guaranteed or confirmed (post 1, Proposition 1.11). Its closed level owes its fidelity to the fixed alphabet; the function has no fixed alphabet and no recipe.
Introduced in: post 1
Also used in: introduction, post 0
31.4 D
31.4.1 Design
An abstract constructor K for a task T is a design to the degree that three conditions hold (post I.5, Definition I.5.0). K is knowledge, so it is resilient. Alternatives to K are possible within one step. And these alternatives perform T much worse, so K is fine-tuned (Marletto 2015, §3.1.1). Fine-tuning is a degree: the larger the share of one-step variants that perform much worse, the more designed K is. In biology the degree is measured by Kimura’s constraint \(1 - k/\mu\), so a weakly constrained site is weakly designed. Design is information plus fine-tuning. The definition mentions no process, so which processes produce designs is a substantive question. A snowflake has order but no design.
Introduced in: post I.5
Lexicon: knowledge condition (D1), alternatives condition (D2), fine-tuning condition (D3), design and adaptive process (F13)
Also used in: introduction, post 7
31.4.2 Drift
Change in a population that no selection drives. Post III.2 takes drift as the null hypothesis for design: a site that is not under selection changes at the rate it mutates, \(k = \mu\). How far a site falls below that, its constraint \(1 - k/\mu\), is the degree to which it is fine-tuned. Drift fails the selection condition, so it is not an adaptive process.
Introduced in: post III.2
Lexicon: drift as null hypothesis (AD3)
Also used in: post III.3
31.5 E
31.5.1 Evolution by natural selection
In this book: the biological instance of the adaptive synthesis task. The task is to stay instantiated in an environment, the medium is the genome, the variation is mutation, recombination and duplication, the elimination is differential death and reproduction, and the resource is food, space, mates and a finite number of births (post III.1). Evolution and the growth of knowledge are two instances of one kind of process.
Introduced in: post III.1
Lexicon: design and adaptive process in biology (AD1)
31.5.2 Explanation
An account of what is there and how it works, as opposed to a mere prediction (Deutsch 2011). Good explanations are hard to vary and have reach: they explain more than the problem they were invented for.
Introduced in: introduction
31.6 F
31.6.1 Fallibilism
No idea is ever fixed as the right one: every idea stays open to criticism, whether or not it happens to be exactly right (post I.6, Definition I.6.12). In this book fallibilism is a property of a process, not of the accuracy of a copy: exact copies occur, but none is treated as final. Formally, a fallibilist process is reflexive and persistently self-reaching in its patterns and criteria. Alone it separates nothing and, without an external anchor, it loses what it knows. The book conjectures that creativity is persistence plus anchored criticism.
Introduced in: post I.6
Lexicon: fallibilism (C10)
31.6.2 Fidelity
How much of an original a copy keeps: the weighted share of the original’s tasks that the copy performs (post I.3, Definition I.3.7). A copy is faithful when its fidelity is 1. Fidelity measures loss only, so a faithful copy can still add tasks of its own. Post 1 applies it to memes, where the added tasks are the gain ε⁺. Sperber’s objection to memetics is that cultural copying has too little fidelity for selection to act on. It holds for culture memes, which the receiver rebuilds. It does not hold for ape memes, which are digital and copied arbitrarily accurately.
Introduced in: post I.3
Lexicon: copy task and ε (F3)
31.6.3 Fixed-criterion process
An adaptive process whose criterion of elimination is fixed and not itself varied, as opposed to a reflexive one (post I.6, Definition I.6.1). Natural selection, with the environment as criterion, is one. A fixed criterion keeps only its own maxima (Proposition I.6.2(a)).
Introduced in: post I.6
Lexicon: reflexivity (C8)
31.7 H
31.7.1 Hard to vary
A good explanation is hard to vary: almost every change to its details would make it stop explaining what it explains (Deutsch 2011, ch. 1). The axial-tilt theory of the seasons is hard to vary, and a myth about gods is not. Post 5 reads hard-to-vary ideas as small points in the space of ideas.
Introduced in: introduction, post 5
Lexicon: convergence (F10)
31.7.2 Host
A system that carries memes, such as a brain. Post 2 models a host as a set of finite states, a computable execution, and two reachability relations: all reachable states, and those reached by the host’s own steps (Definition 2.1).
Introduced in: post 2
31.8 I
31.8.1 Induction
The supposed generalisation from observations to laws. Popper argued that knowledge does not come from induction: no number of white swans proves that all swans are white, and observations do not come with their interpretation attached. Because there is no induction, no copy of a culture meme can be certified exact from the evidence, even when it is exact, and so none is treated as final (fallibilism).
Introduced in: introduction
31.8.2 Information variable, information medium
A set of attributes is an information variable if permuting its members and copying them are possible tasks. A substrate whose attributes form an information variable is an information medium. The interoperability principle says that two information media combine into one, so information can be copied between physically different substrates (Deutsch & Marletto 2015). Information is therefore abstract, since it is not tied to one substrate, and physical, since it is defined by tasks on substrates.
Introduced in: introduction, post 1
31.8.3 Interpreter
A program in one machine’s code that makes the machine perform the tasks of another machine, whose programs are written in a short code (post I.1, §1.2). It is Turing’s universal machine simulating a given one. Constructor theory separates this from building an instance of the other machine and passing it the inputs (Deutsch 2013, §3.7).
Introduced in: post I.1
31.9 K
31.9.1 Knowledge
Information that, once it is instantiated in a suitable environment, tends to cause itself to remain instantiated (Deutsch 2013, §2.15). It is “the most important kind of abstract constructor” (ibid.). The notion is Popper’s knowledge without a knowing subject (Popper 1972, ch. 3): it needs nobody who knows it, so genes contain knowledge as much as theories do. Marletto (2015, §3.1) identifies the information in a recipe with knowledge. In this book knowledge is the first condition of design (the knowledge condition): a design must be resilient, so that it lasts and does not merely occur. Knowledge is not justified true belief, and it is never certain. It grows by conjecture and refutation.
Introduced in: introduction, post 1, post I.5
Lexicon: knowledge condition (D1)
31.10 M
31.10.1 Meme
A unit of cultural transmission (Dawkins 1976, ch. 11). In this book a meme is an abstract constructor: the function that a carrier can perform, fixed by a repertoire of tasks such as explaining, predicting, doing, justifying and passing on. A brain carries the meme if its repertoire contains these tasks (post 1, Definition 1.7). The definition says what a meme is without naming a neural code. A culture meme is not passed from head to head: each receiver rebuilds it from evidence, in its own concepts. An ape meme is a meme whose repertoire is fixed by a string over a fixed behaviour alphabet, and it is copied digitally.
Introduced in: introduction, post 1
Lexicon: meme as equivalence class (F2)
31.10.2 Meme machine
Blackmore’s (1999) name for a human being: a brain that imitates, and so copies memes, and whose evolution memes have driven. Blackmore took the coevolution of genes and memes seriously. This book follows Deutsch (2011, chs. 15–16) instead, whose account rests on Popper’s theory of knowledge and makes no assumption about how memes are implemented in brains.
Introduced in: introduction
31.10.3 Mutation
A change to a variant. Post I.4 follows Chaitin (2012b): a mutation is a program that turns a variant K into K′, and it is drawn with probability \(2^{-n}\) for a program of length \(n\). The mutation distance between two variants is then roughly the length of the shortest program that turns one into the other, not the number of positions in which they differ. In biology, point mutations define the step relation of a genome.
Introduced in: post I.4
Lexicon: algorithmic variation (C6)
31.11 O
31.11.1 Open and closed
A system is open if it is self-reaching and closed otherwise (post I.2, Definition I.2.7; post 2 reads it on a host’s own steps). An ape with fixed behaviour programs, or a system whose ideas are the derivations of a fixed grammar, is closed; its memes are digital strings over a behaviour alphabet. An idle laptop with a compiler is closed too. The distinction is not the line between persons and the rest: a laptop running an enumerator is open, and no condition on reachable sets separates it from a person (Proposition I.2.14).
Introduced in: post I.2
Lexicon: open and closed hosts (F6)
31.11.2 Organism
A living self-reproducer. Its genome is the recipe, and the rest of it is the vehicle, rebuilt from the genome in every generation (Marletto 2015). Dawkins (1976) says that an organism is the vehicle of its genes. Part III reads organisms as the products of an adaptive process, evolution by natural selection.
Introduced in: introduction
31.12 P
31.12.1 Pattern
An abstract constructor that is instantiated in the information variables of a computing system and performs tasks on the system itself (post I.1, Definition I.1.2). A pattern is both constructor and substrate. A meme is a pattern of a brain.
Introduced in: post I.1
31.12.2 Persistence
A system is persistently self-reaching if every state it reaches by its own steps is itself self-reaching (post I.2, Definition I.2.6). In plain words: no state it can get into closes off the alternatives, so nothing it holds is ever final. This is the structure of fallibilism (post I.6). An enumerator is persistent too, so persistence alone does not make a person.
Introduced in: post I.2
Lexicon: self-reachability (C3)
Also used in: post I.0, post 1, post 3, post III.3
31.12.3 Possible and impossible
A task \(X\) is possible, written \(X^{\checkmark}\), if the laws of physics impose no bound, short of perfection, on the accuracy and the reliability with which it can be performed. Otherwise it is impossible, written \(X^{\text{✘}}\) (Deutsch 2013, §1.1; Deutsch & Marletto 2025, §3, eqs. 2–3). Possible does not mean exact: no perfect constructor exists. It does not mean that the task is ever performed either. The book’s ε-form: for every ε > 0 there is an approximate constructor whose inaccuracy and unreliability are both at most ε.
Introduced in: introduction, post 1
Also used in: post 0, post III.4
31.12.4 Programmable constructor
A constructor whose behaviour is set by a program it holds, as a ribosome’s product is set by the RNA it reads (Deutsch 2013, §2.1). In Marletto’s architecture of a self-reproducer it is the part P of the vehicle that executes the recipe.
31.13 R
31.13.1 Rational meme
A meme that is “created by rational and critical thought” and depends “on such thought to get [itself] replicated faithfully” (Deutsch 2011, ch. 15). In this book a rational meme is one whose criterion of copying is anchored primarily outside the meme and its hosts, in reality (post 4).
Introduced in: introduction
31.13.2 Recipe
The digital information that a self-reproducer copies and executes, R in Marletto’s (2015) architecture. It is the replicator. The recipe must be digital, so that errors can be corrected, and modular, so that elementary constructors can execute it step by step. Marletto calls the information in the recipe knowledge. For a gene the recipe is the base sequence. Among memes only the closed level of copying has recipe-like strings: strings over a closed alphabet that is fixed in the receiver, as in the steps of an ape meme, phonemes and words, writing, recitation, checklists and forks of source code. Most of them are copied by reading the string back from a performance, not from a separate recipe, and their fidelity comes from the fixed alphabet and redundancy. A recipe is copied from a recipe, as in genes, only where a text is copied from the text. The function of a culture meme, which the receiver rebuilds from reasons, is not a recipe (post 6).
Introduced in: introduction, post 1
Lexicon: recipe–vehicle architecture (F11)
Also used in: post 0
31.13.3 Reflexivity
An adaptive process is reflexive if its criterion of elimination is itself carried in the medium, so that it is copied, varied and eliminated like other patterns (post I.6, Definition I.6.1). Science is reflexive: what counts as a refutation is a conjecture too.
Introduced in: post I.6
Lexicon: reflexivity (C8)
31.13.4 Reliability
How well a device keeps its ability to perform a task to a given accuracy “when re-used repeatedly” (Deutsch & Marletto 2025, §2, following Violaris & Marletto 2022). Accuracy is about one run. Reliability is about the device staying able to repeat it. Deutsch (2013, §1.1) says the same in other words: “how well things that are capable of approximately performing it could retain their ability to do so”.
Introduced in: introduction, post 1
31.13.5 Repertoire
The set of tasks, from a fixed task universe, that a program or a brain state performs (post I.1, Definition I.1.6; post 1, Definition 1.6). A meme is fixed by a repertoire, and a brain carries the meme if the brain’s repertoire contains it.
31.13.6 Replicator
Anything that gets copied and whose copies get copied in turn (Dawkins 1976). Deutsch (2011, ch. 4) puts it as “anything that contributes causally to its own copying”. Genes are the replicators of biology. In Deutsch’s sense a culture meme is a replicator: it causes its holders to explain and pass it on. In Marletto’s (2015) architecture the replicator is the recipe, and in that sense only strings on the closed level, such as ape memes and the words that carry a culture meme, are replicators; the function of a culture meme has no recipe, because each receiver rebuilds it from reasons.
Introduced in: introduction
31.14 S
31.14.1 Self-reaching, self-reachability
A computing system is self-reaching if every partial computable function is the behaviour of some state that it reaches by its own steps, without programming from outside (post I.2, Definition I.2.6). It is the book’s formal reading of a universal explainer, and of an open host. It is necessary for a person, not sufficient: a laptop running a program enumerator is self-reaching too.
Introduced in: post I.2
Lexicon: self-reachability (C3)
31.14.2 Self-reproducer
A constructor that builds a copy of itself from generic resources. Marletto (2015) shows that accurate self-reproduction is possible under laws that contain no design, provided that the laws allow digital information media. The self-reproducer must then consist of a recipe (the replicator) and a vehicle, and the vehicle of a copier and a programmable constructor, as von Neumann (1966) described.
Introduced in: post 1
Lexicon: recipe–vehicle architecture (F11)
31.14.3 Serial and parallel composition
The serial composition \(AB\) of tasks performs \(B\) and then \(A\) on the same substrate; the parallel composition \(A \otimes B\) performs \(A\) on one substrate and \(B\) on another, which together form the composite system \(\mathcal{P}_1 \oplus \mathcal{P}_2\) (Deutsch 2013, §1.2). The composition principle says that every network of possible tasks, with outputs fitting inputs, is possible. A possible task can still be made of impossible ones.
Introduced in: introduction
31.14.4 Short code and complete code
Von Neumann’s names for a programmer’s language and a machine language, the machine’s secondary and primary language (von Neumann 1958; 1966, editor’s introduction). A program in complete code runs on the circuit directly; a program in short code runs through an interpreter (post I.1, §1.2).
Introduced in: post I.1
31.14.5 Software
An abstract constructor running on a programmable constructor (post I.1, §1.2). It is substrate-independent because each layer of a software stack needs only that the layer below performs its tasks, not how.
Introduced in: post I.1
31.14.6 Static society
A society “changing on a timescale unnoticed by the inhabitants” (Deutsch 2011, ch. 15). Deutsch argues that anti-rational memes dominate such societies and suppress change. Post 4 models the stasis.
Introduced in: introduction
31.14.7 Step relation
The relation that holds between two states if the system takes one to the other in one step (post I.1, Definition I.1.3; post 3, Definition 3.1). For variants of a constructor, a step relation is the elementary variation: a point mutation for genes, one conjectural step for ideas (post I.5).
31.14.8 Substrate
A physical system that a task transforms (Deutsch 2013). Two substrates that are jointly the input of a task form a composite substrate, written with ⊕.
Introduced in: introduction, post 1
Also used in: post I.0
31.14.9 Synthesis
The task that takes a specification and an empty program medium to the same specification and a program that meets it (post I.3, Definition I.3.2). Its product is an abstract constructor, so a synthesiser is a constructor that builds constructors. Copying a culture meme is synthesis from examples.
Introduced in: post I.3
Lexicon: synthesis (C4)
31.15 T
31.15.1 Task
A specification of a transformation: an input attribute \(x\) and an output attribute \(y\) of a substrate, written \(x \to y\) on \(\mathcal{P}\) (Deutsch & Marletto 2025, eq. 1), or a set of such pairs (Deutsch 2013). A task says what is to be done, not how, and not when.
Introduced in: introduction, post 1
Also used in: post I.0
31.15.2 Timer
A constructor whose output says whether a given period has elapsed since it was started (Deutsch & Marletto 2025, §5–6). Timers let a theory without a time parameter speak of duration, and so of how quickly a task is done. Post I.1 builds one as a Push program, and post 3 uses it to define t-Reachable.
Introduced in: introduction, post 3
Lexicon: time and efficiency (F5)
31.15.3 Transpose
The transpose \(X^{\sim}\) of a task \(X\) swaps its inputs and outputs (Deutsch 2013, §1.1). A possible task can have an impossible transpose: converting work entirely into heat is possible, the reverse is not. This is how constructor theory states irreversibility without a time parameter.
Introduced in: introduction
31.16 U
31.16.1 Universal computer
A system that can be brought to perform every computable behaviour by programming from outside (post I.2, Definition I.2.6): a universal Turing machine, and the universal constructor restricted to computations. A laptop is one, given enough memory. A universal explainer must reach every behaviour by its own steps.
Introduced in: post I.2
31.16.2 Universal constructor
A programmable constructor whose repertoire, given the right program, contains every task the laws of physics allow (Deutsch 2013; post I.2). It is an abstract machine, read “given enough resources”. Restricted to computations it is a universal computer. A laptop is a universal computer but not a universal constructor, because it cannot be programmed to build a chair. Deutsch (2013, §3.15) doubts that people are universal constructors, because they cannot be programmed.
Introduced in: post I.2
31.16.3 Universal explainer
Deutsch (2011, chs. 3 and 6) argues that people are universal explainers: anything that can be understood at all can in principle be understood by a person. Post I.2 states a formal version, self-reachability: every computable behaviour lies among the states the system reaches by its own steps. It proves that this separates a person from an idle laptop, but not from a laptop running a program enumerator.
Introduced in: introduction, post I.2
Lexicon: self-reachability (C3)
31.17 V
31.17.1 Variation
The production of new variants from existing ones. In an adaptive synthesis task it is done by a variation constructor V, which must not read the task (the unspecific-variation condition). Variation may be structured: it may recombine, duplicate and reuse parts. Algorithmic variation measures the size of a variation by the description length of the change (post I.4).
Introduced in: post I.4, post I.5
Lexicon: algorithmic variation (C6), unspecific-variation condition (A1′)
31.17.2 Vehicle
The part of a self-reproducer that does the work, V in Marletto’s (2015) architecture: a copier C that copies the recipe and a programmable constructor P that executes it. The vehicle is built afresh from the recipe in every generation, so errors in the vehicle are not inherited. Only errors in the recipe are, and they are what natural selection acts on. An organism is a vehicle together with the genome that it carries. The word comes from Dawkins (1976), for whom an organism is the vehicle of its genes.
Introduced in: introduction, post 1
Lexicon: recipe–vehicle architecture (F11)
Also used in: post 0
31.17.3 Virtual machine
A programmable constructor that is itself an abstract constructor: it exists only as a program running on another machine, and its programs are information in that machine’s memory (post I.1, §1.2). The Push3 interpreter, running on the JVM, is the virtual machine of Part I.
Introduced in: post I.1
31.18 W
31.18.1 World 3
Popper’s world of the objective contents of thought: theories, problems, arguments and their logical relations, next to the physical World 1 and the mental World 2 (Popper 1972, ch. 3). A theory in World 3 is written down, so others can criticise it; the book takes this as the conjecture that criticism needs a representation (post I.6, Conjecture I.6.15). Deutsch and Marletto’s knowledge without a knowing subject comes from here.
Introduced in: introduction
source: book2/notebooks/meme_constructor/posts/glossary.clj