Essay II

Operadic Autopoiesis in an (∞)-Topos

A cell is a requirement operad inside a least fixed point of endogenous production. A virus closes only after a homotopy pushout with a host. A computer, on the physical signature, does not close at all.

Contents60 sections

Abstract

Let $E$ be an $(\infty)$ -topos containing an internal symmetric monoidal $(\infty)$ -category $P^{\otimes}$ of open physical processes, and let $O^{\otimes}$ be the associated internal colored $(\infty)$ -operad. Colors are interfaces, resources, and organizational roles. Multimorphisms are composable physical operations.

For a system $X$ , write $E_{X}^{\otimes} ↪ O^{\otimes}$ for the endogenous process operad, $R_{X}^{\otimes} ↪ O^{\otimes}$ for the required organizational operad, and $G_{X}^{\otimes}$ for the least fixed point of an accessible production operator applied to $E_{X}$ . Operadic autopoiesis is the inclusion

R_{X}^{\otimes} \subseteq G_{X}^{\otimes} .

Every organizational operation required for the continuation of $X$ is realizable by a recursively self-generated network internal to $X$ .

Interaction across a common interface operad $I^{\otimes}$ is the homotopy pushout $E_{X}^{\otimes} ⨿_{I} E_{H}^{\otimes}$ . The pushout contains mixed composites, not merely the union of the two operads. A host-dependent replicator then satisfies $R_{V} ⊈ G_{V}$ but $R_{V} \subseteq G_{V ⋆ H}$ . A cell satisfies $R_{B} \subseteq G_{B}$ . An ordinary computer, once fabrication, repair, replacement, and physical reproduction are required, satisfies $R_{C} ⊈ G_{C}$ .

These properties are invariant under equivalence, local-to-global along effective epimorphisms, and separate autonomous systems from externally dependent ones. A non-conservativity theorem explains why equivalence of dependency patterns, or of computational observations, does not yield equivalence of physical systems.

1Ambient (∞)-topos

Let $E$ be an $(\infty)$ -topos, interpreted as a generalized space of physical contexts: spacetime regions, experimental conditions, environments, or organizational scales. Systems and processes may vary locally, and they satisfy descent. Internal higher categories in an $(\infty)$ -topos may be studied as suitable sheaves of $(\infty)$ -categories. The internal theory required below is available in that setting.

2Internal (∞)-categories

Write $Cat_{\infty} (E)$ for the $(\infty)$ -category of internal $(\infty)$ -categories in $E$ . Concretely, an internal $(\infty)$ -category may be presented as a complete Segal object $C_{∙} : Δ^{op} \to E$ . Then $C_{0}$ is the object of objects, $C_{1}$ the object of morphisms, and $C_{n}$ the object of coherently composable $n$ -fold strings. The Segal equivalences read

C_{n} ≃ C_{1} \times_{C_{0}} \dots \times_{C_{0}} C_{1},

and completeness identifies equivalences internally.

3Internal symmetric monoidal (∞)-categories

Definition 3.1

Internal symmetric monoidal (∞)-category

An internal symmetric monoidal $(\infty)$ -category is a commutative algebra

P^{\otimes} \in CAlg (Cat_{\infty} (E)) .

The multiplication $\otimes : P \times P \to P$ is parallel composition. The categorical composition $\circ$ is sequential composition. The symmetric monoidal coherences hold internally in $E$ .

4Open physical processes

Interpret $P^{\otimes}$ as the internal symmetric monoidal $(\infty)$ -category of open physical processes. Objects are interfaces: chemical species, membranes, material stocks, energy carriers, molecular machines, electrical ports, mechanical ports. A morphism $p : A \to B$ is an open process taking the input interface $A$ to the output interface $B$ .

Definition 4.1

Open realization

An open process may be represented by a cospan-shaped diagram $L (A) \to X \leftarrow L (B)$ , where $X$ is the system performing the transformation and $L (A)$ , $L (B)$ are its exposed interfaces. Sequential composition is gluing along a common interface, $X ⨿_{L (B)} Y$ . Parallel composition is $X \otimes Y$ .

This is the higher-categorical analogue of the structured cospan used for open reaction networks, electrical circuits, and Petri nets. Ordinary pushouts are replaced by homotopy-coherent pushouts.

5The process (∞)-operad

A symmetric monoidal $(\infty)$ -category determines an $(\infty)$ -operad. In Lurie’s formalism an $(\infty)$ -operad is a map $O^{\otimes} \to N (Fin_{*})$ with inert lifts and Segal conditions. We use the internal analogue.

Definition 5.1

Process operad

Let $O^{\otimes}$ be the internal $(\infty)$ -operad underlying $P^{\otimes}$ . For colors $x_{1}, \dots, x_{n}, y$ there is an object

Mul_{O} (x_{1}, \dots, x_{n}; y) \in E

of physical $n$ -ary operations $(x_{1}, \dots, x_{n}) \to y$ . Operadic substitution supplies

Mul (y_{1}, \dots, y_{n}; z) \times i \prod Mul (x_{i 1}, \dots, x_{i k_{i}}; y_{i}) ⟶ Mul (x_{11}, \dots; z) .

A physical causal network is a tree of composable operations, not merely a sequence of maps.

6Internal suboperads

Definition 6.1

Suboperad

A suboperad $A^{\otimes} ↪ O^{\otimes}$ consists of subobjects

Mul_{A} (x; y) ↪ Mul_{O} (x; y)

for every profile $(x; y)$ , stable under units, permutations, inert operations, operadic substitution, and restriction in $E$ . Write $SubOp (O)$ for the poset of suboperads, ordered by inclusion. We write both $A \leq B$ and $A \subseteq B$ for that order.

Lemma 6.2

Intersections of suboperads

An arbitrary intersection of suboperads of $O^{\otimes}$ is a suboperad.

Proof. Intersection is computed profilewise,

Mul_{⋂_{i} A_{i}} (x; y) = i ⋂ Mul_{A_{i}} (x; y) .

Each factor contains the operadic units and is stable under permutation and substitution, so the intersection is as well.

End of proof.

Definition 6.3

Generated suboperad

For a collection $S$ of operations of $O$ , the generated suboperad $⟨ S ⟩_{O}$ is the intersection of all suboperads containing $S$ . It is the smallest suboperad containing $S$ .

7Systems as bounded operadic realizers

Definition 7.1

Structured physical system

A structured physical system is a tuple

X = (∣ X ∣, \partial X, E_{X}^{\otimes}, R_{X}^{\otimes}, Φ_{X}) .

Here $∣ X ∣ \in E$ is the physical carrier, $\partial X$ its interaction boundary, $E_{X}^{\otimes} ↪ O^{\otimes}$ its primitive endogenous process operad, $R_{X}^{\otimes} ↪ O^{\otimes}$ its organizational requirement operad, and $Φ_{X}$ its recursive production operator.

8Endogenous processes

Definition 8.1

Endogenous operation

An operation $f \in Mul_{O} (x; y)$ is endogenous to $X$ when a physical realizer of $f$ lies inside the causal boundary of $X$ . The endogenous suboperad $E_{X}^{\otimes}$ contains every operation realizable from structures belonging to $X$ , and it is closed under operadic composition: if $f, g_{1}, \dots, g_{n} \in E_{X}$ , then $f \circ (g_{1}, \dots, g_{n}) \in E_{X}$ .

9Organizational requirements

Endogeneity is not enough. A rock has endogenous physical processes. The question is whether the processes required for the organization to persist are generated by that organization.

Definition 9.1

Requirement operad

The requirement operad $R_{X}^{\otimes}$ is the suboperad generated by those organizational operations whose continued realization is required for $X$ to remain in the chosen viability class. Generators may include boundary maintenance, repair, catalysis, translation, fabrication, replication, regulation, and energy transduction. The signature is selectable. Different theories of life correspond, in part, to different choices of $R_{X}^{\otimes}$ .

10Why one-step generation is not enough

Suppose an organism currently possesses a ribosome. Translation is then an endogenous operation. That does not say whether the system can produce or repair the machinery that realizes translation. Closure has to be recursive, not a snapshot of the operations available at one moment.

11Production doctrine

Definition 11.1

Production doctrine

A production doctrine for $X$ is a monotone endomap $Φ_{X} : SubOp (O) \to SubOp (O)$ . The value $Φ_{X} (A)$ contains those organizational operations whose physical realizers $X$ can construct, maintain, regenerate, or repair using operations from $A$ , together with the primitive endogenous capabilities and the permitted matter–energy inputs. Monotonicity means $A \leq B$ implies $Φ_{X} (A) \leq Φ_{X} (B)$ .

Assumption 11.2

Accessibility

The operator $Φ_{X}$ is accessible, and inflationary once the primitive endogenous operations have been adjoined. For a sufficiently large regular ordinal $λ$ , $Φ_{X}$ preserves the relevant $λ$ -filtered colimits. Recursive generation may therefore be computed by transfinite iteration.

12Recursive endogenous closure

Set $G_{X}^{(0)} := E_{X}$ . At a successor ordinal,

G_{X}^{(α + 1)} := Φ_{X} (G_{X}^{(α)}) .

At a limit ordinal $β$ ,

G_{X}^{(β)} := ⟨ α < β ⋃ G_{X}^{(α)} ⟩ .

For sufficiently large $λ$ the iteration stabilizes. Write $G_{X} := G_{X}^{(λ)}$ , and also $G_{X} = μ Φ_{X}$ for this least fixed point.

Proposition 12.1

Fixed point

The operad $G_{X}$ is a fixed point of $Φ_{X}$ : $Φ_{X} (G_{X}) ≃ G_{X}$ .

Proof. By accessibility,

Φ_{X} (α < λ colim G_{X}^{(α)}) ≃ α < λ colim Φ_{X} (G_{X}^{(α)}) .

The right-hand side is the colimit of the successor stages $G_{X}^{(α + 1)}$ . Shifting the ordinal index does not change the colimit, so $Φ_{X} (G_{X}) ≃ G_{X}$ .

End of proof.

Proposition 12.2

Least fixed point

The operad $G_{X}$ is the least $Φ_{X}$ -closed suboperad containing $E_{X}$ .

Proof. Let $A$ contain $E_{X}$ and satisfy $Φ_{X} (A) \leq A$ . Transfinite induction gives $G_{X}^{(α)} \leq A$ for every $α$ . Successor stages use monotonicity. Limit stages use that $A$ is itself a suboperad containing every earlier stage, hence contains the suboperad they generate. Therefore $G_{X} \leq A$ .

End of proof.

13Operadic organizational closure

Definition 13.1

Operadic autopoiesis

The system $X$ is organizationally closed, written $Aut (X)$ , when

R_{X}^{\otimes} \subseteq G_{X}^{\otimes} .

Every operation organizationally required by $X$ lies in the recursively generated endogenous operad.

14Interpretation

The inclusion $R_{X} \subseteq G_{X}$ does not say that $X$ is materially isolated. Environmental resource morphisms $M_{env} \to X$ and $E_{env} \to X$ are allowed. Matter and energy may flow through the system. Closure concerns the generation of organizational constraints and of productive machinery.

This matches the classical autopoietic idea: a living system recursively produces the network and the boundary that constitute it.

15Equivalence of structured systems

Definition 15.1

Equivalence

An equivalence $X ≃ Y$ of structured systems consists of compatible equivalences $∣ X ∣ ≃ ∣ Y ∣$ , $\partial X ≃ \partial Y$ , $E_{X} ≃ E_{Y}$ , and $R_{X} ≃ R_{Y}$ , conjugating the production operators:

F_{*} \circ Φ_{X} ≃ Φ_{Y} \circ F_{*} .

16Fixed-point invariance

Theorem 16.1

Equivalences preserve generative closure

If $X ≃ Y$ is an equivalence of structured systems, then $F_{*} (G_{X}) ≃ G_{Y}$ .

Proof. At stage zero, $F_{*} (E_{X}) ≃ E_{Y}$ , so $F_{*} (G_{X}^{(0)}) ≃ G_{Y}^{(0)}$ . If the claim holds at stage $α$ , conjugacy of the production operators gives

F_{*} Φ_{X} (G_{X}^{(α)}) ≃ Φ_{Y} F_{*} (G_{X}^{(α)}) ≃ Φ_{Y} (G_{Y}^{(α)}),

which is the successor stage of $Y$ . At limit ordinals, $F_{*}$ preserves the generated colimits in question. By transfinite induction the stages agree, and therefore so do the stable values.

End of proof.

17Autopoiesis is invariant

Theorem 17.1

Invariance of operadic autopoiesis

If $X ≃ Y$ , then $Aut (X)$ if and only if $Aut (Y)$ .

Proof. Suppose $R_{X} \subseteq G_{X}$ . Applying $F_{*}$ and using preservation of requirements together with Theorem 16.1 yields $R_{Y} \subseteq G_{Y}$ . The converse uses $F^{- 1}$ .

End of proof.

Corollary 17.2

Separation by autopoiesis

If $Aut (X)$ and $\neg Aut (Y)$ , then $X \neq ≃ Y$ .

This already separates an autonomously closed cell from an externally maintained machine, inside the present model.

18Interaction interfaces

Let $X$ and $H$ interact through an interface operad $I^{\otimes}$ , with operad morphisms $i_{X} : I^{\otimes} \to E_{X}^{\otimes}$ and $i_{H} : I^{\otimes} \to E_{H}^{\otimes}$ . Interface colors may include membrane receptors, ribosomal binding sites, ATP, nucleotides, polymerases, and transport machinery.

19Operadic coupling

Definition 19.1

Coupled operad

The raw coupled operad is the homotopy pushout

E_{X} ⨿_{I} E_{H} .

Whenever the two endogenous realizations agree on the common interface, there is a canonical morphism

j : E_{X} ⨿_{I} E_{H} ⟶ O .

Write $E_{X ⋆ H} := ⟨ im (j) ⟩_{O}$ for the suboperad it generates.

20Why the pushout is larger than the union

The pushout contains more than $E_{X} \cup E_{H}$ . It contains operations obtained by composing an operation from one system with operations from the other. Schematically, $f_{X} \circ (g_{H}, h_{X}, k_{H})$ is legal whenever the colors match. The coupled system has mixed causal pathways that neither system has alone.

21Mixed trees

Lemma 21.1

Mixed-tree lemma

Every finite operadic tree whose vertices are labeled by operations from $E_{X}$ or from $E_{H}$ , and whose cross-system internal edges have colors in the common interface $I$ , determines an operation of $E_{X ⋆ H}$ .

Proof. Induct on height. A tree of height zero is a single operation of $X$ or of $H$ , hence lies in the image of the pushout. If the claim holds up to height $n$ , and $T$ has height $n + 1$ , each immediate subtree of the root represents, by induction, an operation of the pushout-generated operad. The root operation lies in $E_{X}$ or in $E_{H}$ . Matching edge colors make the operadic substitution defined, and closure under substitution produces the operation corresponding to $T$ .

End of proof.

22Emergent causal capacity

Corollary 22.1

Coupling creates operations

An operation $r$ may satisfy $r \in / E_{X}$ and $r \in / E_{H}$ while $r \in E_{X ⋆ H}$ .

Proof. The operation $r$ may be a composite tree with some vertices supplied by $X$ and others by $H$ . Lemma 21.1 places that tree in $E_{X ⋆ H}$ .

End of proof.

This is the structure needed for viral use of host-cell machinery.

23Coupled production

The coupled system has a production doctrine $Φ_{X ⋆ H}$ whose primitive stage contains $E_{X ⋆ H}$ . Its recursively generated operad is $G_{X ⋆ H} := μ Φ_{X ⋆ H}$ .

24Relative closure

Definition 24.1

Closure relative to a host

The system $X$ is organizationally closed relative to $H$ , written $Aut (X ∣ H)$ , when $R_{X} \subseteq G_{X ⋆ H}$ .

25Strict host dependence

Definition 25.1

Strict dependence

The system $X$ is strictly dependent on $H$ when $\neg Aut (X)$ but $Aut (X ∣ H)$ . Equivalently, $R_{X} ⊈ G_{X}$ while $R_{X} \subseteq G_{X ⋆ H}$ .

26The virus theorem

Let $V$ be a virus and $H$ a susceptible host cell. Viral replication depends on host translation, and many viruses co-opt host proteins, membranes, metabolism, and other cellular processes for genome replication and assembly. The modeling hypothesis is

R_{V} ⊈ G_{V}, R_{V} \subseteq G_{V ⋆ H} .

Theorem 26.1

Viral host-dependence

Under these hypotheses, $V$ is a strictly host-dependent organizational system.

Proof. This is Definition 25.1. The content is in the structure that instantiates the predicates: $G_{V ⋆ H}$ contains mixed virus–host operations produced by the operadic pushout, which $G_{V}$ does not.

End of proof.

27The cell

Let $B$ be a bacterial cell, and let $R_{B}$ be generated by membrane maintenance, repair, translation, replication, metabolism, regulation, and the other modes treated as essential. The biological-autonomy hypothesis is $R_{B} \subseteq G_{B}$ . This is an axiomatization, not a theorem of biology. Inside the theory it says that the cell’s required organizational machinery lies in its recursively regenerative operadic closure.

28Virus and cell

Theorem 28.1

Virus–cell separation

If $Aut (B)$ and $\neg Aut (V)$ , then $B \neq ≃ V$ as structured physical systems.

Proof. Operadic autopoiesis is invariant under equivalence by Theorem 17.1. An equivalence $B ≃ V$ would force $Aut (B)$ if and only if $Aut (V)$ .

End of proof.

29The computer

Let $C$ be a conventional computer. On a restricted operational signature — instruction execution, memory access, arithmetic — the machine may have substantial endogenous closure. Take instead the physically complete requirement operad $R_{C}^{phys}$ generated by chip fabrication, replacement, repair, structural reproduction, power infrastructure, and assembly. For an ordinary present-day computer,

R_{C}^{phys} ⊈ G_{C},

while an industrial infrastructure $I$ supplies the missing operations, $R_{C}^{phys} \subseteq G_{C ⋆ I}$ . Under the physical requirement signature, $\neg Aut (C)$ .

30Cell and computer

Theorem 30.1

Cell–computer separation

If $Aut (B)$ and $\neg Aut (C)$ , then $B \neq ≃ C$ as operadically structured physical systems.

Proof. Theorem 17.1.

End of proof.

31Stronger than a list of parts

A purely material argument says $Supp (B) \neq ≃ Supp (C)$ . The operadic argument says $G_{B} \neq ≃ G_{C}$ as organizational process algebras. The first is a difference of objects. The second is a difference of the higher algebra of transformations among objects.

32Material typing

The substrate invariant remains available. Let $K \in E$ be the object of material component types, and let $τ_{X} : C_{X} \to K$ be the typing of the components of $X$ . Set $Supp (X) := im (τ_{X})$ .

Proposition 32.1

Support is invariant

An equivalence that preserves material typing induces $Supp (X) ≃ Supp (Y)$ .

Proof. As in the typed setting: if $τ_{X} ≃ τ_{Y} \circ f$ with $f$ an equivalence, the images agree.

End of proof.

Material support is an invariant independent of the closure truth value.

33The full physical invariant

Define

J (X) = (Supp (X), E_{X}, R_{X}, Φ_{X}, G_{X}) .

Theorem 33.1

Structured equivalence preserves $\mathcal{J}$

A structured equivalence $X ≃ Y$ induces $J (X) ≃ J (Y)$ . Any difference among these components obstructs physical equivalence.

Proof. Support is Proposition 32.1. The operads $E$ and $R$ , and the operator $Φ$ , are preserved by Definition 15.1. The generated operad $G$ is preserved by Theorem 16.1.

End of proof.

34Why a virus and a program can look alike

Let $V$ be a virus and $P$ a self-copying program. There is a coarse dependency pattern $V + H \to V^{⊔ n}$ and $P + M \to P^{⊔ n}$ . Define a forgetful functor $Q : PhysSys \to Dep$ that remembers only specification, host, and replication. It is entirely possible that $Q (V) ≃ Q (P)$ .

35Non-conservativity

Theorem 35.1

Forgetting support or the operad destroys conservativity

If $Q$ forgets either material support or the endogenous operad, then $Q$ is not conservative in general. Thus $Q (X) ≃ Q (Y)$ does not imply $X ≃ Y$ .

Proof. Choose $X$ and $Y$ with the same dependency diagram and with $Supp (X) \neq ≃ Supp (Y)$ . Then $Q (X) ≃ Q (Y)$ . If $X ≃ Y$ held in the full category, Proposition 32.1 would force the supports to agree.

End of proof.

36The virus/program analogy

Therefore $Q (V) ≃ Q (P)$ may hold even though $V \neq ≃ P$ . The analogy lives in a quotient of the structural theory. It is invalid to infer $V ≃ P$ from $Q (V) ≃ Q (P)$ , and still more invalid to infer $B ≃ P$ for a bacterium $B$ .

37Localization

The same phenomenon can be phrased as a localization. Let $L : E \to E_{comp}$ be an accessible left-exact localization, read as a computational observational modality. Interpret $L X$ as what is observable about $X$ through the chosen computational semantics.

Definition 37.1

Computational equivalence

Systems $X$ and $Y$ are computationally equivalent when $L X ≃ L Y$ .

Proposition 37.2

Non-conservative localizations identify non-equivalent systems

If $L$ is not conservative, there exist $X \neq ≃ Y$ with $L X ≃ L Y$ .

Proof. Failure of conservativity is exactly failure to reflect equivalences.

End of proof.

38Simulation

Suppose a bacterium and a computer simulation satisfy $L (B) ≃ L (C)$ . All that follows is equality after computational localization. One cannot conclude $B ≃ C$ unless $L$ reflects equivalences on the relevant class of systems.

39Locality

Let $u : U ↠ 1$ be an effective epimorphism, and assume the production doctrine is geometric:

u^{*} Φ_{X} (A) ≃ Φ_{u^{*} X} (u^{*} A) .

Inverse-image functors preserve colimits, so transfinite generation commutes with pullback: $u^{*} G_{X} ≃ G_{u^{*} X}$ .

40Descent for closure

Theorem 40.1

Closure descends

If $u^{*} R_{X} \subseteq u^{*} G_{X}$ for an effective epimorphism $u : U ↠ 1$ , then $R_{X} \subseteq G_{X}$ .

Proof. Suboperad inclusion is profilewise inclusion of subobjects. For each multimorphism profile, $u^{*} R (x; y) \leq u^{*} G (x; y)$ . Pullback along an effective epimorphism reflects subobject inclusion, so $R (x; y) \leq G (x; y)$ for every profile.

End of proof.

41Host coupling is not a change of context

Corollary 41.1

Observation does not create autonomy

If $\neg Aut (V)$ , effective-epimorphic base change alone cannot produce $Aut (u^{*} V)$ .

Proof. If $Aut (u^{*} V)$ held, Theorem 40.1 would give $Aut (V)$ .

End of proof.

The host has to contribute operations. That is the point of the pushout.

42Three loci

Suppose structured systems form an internal moduli object, or a stack, $Sys$ over $E$ . The autonomous locus, the merely dependent locus, and the host-dependent locus are the internally definable subobjects

Aut Dep HostDep = {X ∣ R_{X} \subseteq G_{X}}, = {X ∣ R_{X} ⊈ G_{X}}, = {X ∣ \neg Aut (X) \land \exists H . Aut (X ∣ H)} .

Thus $Aut ↪ Sys$ and $HostDep ↪ Sys$ are internal loci, not external labels pasted onto the mathematics.

43An important non-theorem

It does not follow that $HostDep$ is the class of viruses. Computers can be externally dependent. Parasitic organisms can be externally dependent. Organelles can be externally dependent. Dependence is an invariant. It is not a definition of “virus.” Further invariants are required.

44A stratified invariant

Enrich the invariant by hereditary organization, lineage, and self-produced boundary:

K (X) = (Supp (X), G_{X}, R_{X}, Her (X), Lin (X), Bound (X)) .

Cells, viruses, and computers may then occupy distinct strata even where one closure coordinate agrees.

45One invariant proves little

Suppose $Aut (V)$ and $Aut (C)$ are both false. It does not follow that $V ≃ C$ . Both objects merely lie in the same fiber of $Aut : Sys \to Ω$ . A fiber can contain many non-equivalent objects. The locus $Aut^{- 1} (⊥)$ is, potentially, enormous.

46A hierarchy of forgetful semantics

There is a chain of forgetful functors

PhysSys F_{1} OrgSys F_{2} DepSys F_{3} CompSem .

The first forgets detailed substrate, the second forgets most organizational structure, and the third retains only computationally relevant behavior. One may have

F_{3} F_{2} F_{1} (B) ≃ F_{3} F_{2} F_{1} (C)

while $B \neq ≃ C$ .

47Forgetfulness in general

Theorem 47.1

Reflection is conservativity

Let $F : C \to D$ be any functor. The inference $F (X) ≃ F (Y) ⟹ X ≃ Y$ , for all $X$ and $Y$ , holds if and only if $F$ is conservative.

Proof. A functor is conservative precisely when it reflects equivalences.

End of proof.

Corollary 47.2

The missing step

Any argument that identifies biological and computational systems through a functor forgetting metabolic, material, organizational, boundary, or reproductive structure must separately establish conservativity. Without that step the conclusion does not follow.

48Main separation theorem

Theorem 48.1

Operadic separation

Let $B$ , $V$ , and $C$ be a cellular organism, a virus, and a conventional computer in $Sys_{E, O}$ . Assume $R_{B} \subseteq G_{B}$ ; $R_{V} ⊈ G_{V}$ ; $R_{V} \subseteq G_{V ⋆ H}$ for some host $H$ ; and $R_{C} ⊈ G_{C}$ . Assume also that at least one of $Supp (V)$ , $G_{V}$ , $Her (V)$ , and $Lin (V)$ differs from the corresponding invariant of $C$ . Then

B \neq ≃ V, B \neq ≃ C, V \neq ≃ C

in the fully structured category.

Proof. The first two non-equivalences are Theorem 17.1, since $B$ is organizationally closed and $V$ and $C$ are not. For the third, an equivalence would preserve every component of $K$ . By hypothesis at least one component differs.

End of proof.

49What the equations say about a bacterium

The bacterium is not distinguished merely by having proteins rather than transistors. The stronger property in the model is $R_{B} \subseteq μ Φ_{B}$ : the operations required to regenerate the organism lie in the least recursively closed operadic algebra generated by processes internal to the organism.

50What the equations say about a virus

For a virus, $R_{V} ⊈ μ Φ_{V}$ , but $R_{V} \subseteq μ Φ_{V ⋆ H}$ . Organizational closure of the virus is a property of the coupled virus–host system. The pushout $E_{V} ⨿_{I} E_{H}$ is what supplies the missing mixed operations.

51What the equations say about a computer

For a computer, $R_{C}^{phys} ⊈ μ Φ_{C}$ , while one may have $R_{C}^{phys} \subseteq μ Φ_{C ⋆ I}$ , where $I$ is an industrial-technological system. The analogue of the viral host is not another computer. It is the industrial ecology that manufactures and maintains computers.

52A sharper analogy, still not an equivalence

In one coordinate, the pair (virus, cell) is analogous to the pair (computer, industrial civilization). Both are externally completed systems, $V ⋆ H$ and $C ⋆ I$ . The analogy still does not give $V ≃ C$ . The coupled operads, material types, interfaces, lineages, and organizational constraints remain different.

53Higher homotopies

An $(\infty)$ -operad, rather than an ordinary operad, matters when a biological operation has many realizations. If $r$ can be implemented by processes $f$ and $g$ , one need not impose $f = g$ . There may be a path $f ≃ g$ , and paths between paths. The multimorphism object $Mul_{O} (x; y)$ retains the homotopy type of the space of realizations. Organizational roles can be invariant under coherent replacement of realizers.

54Robustness as homotopy

A biological function may correspond not to one operation but to a connected component $[f] \in π_{0} Mul_{O} (x; y)$ . Robust organization can involve large spaces of mutually substitutable realizations. Functional degeneracy can be probed by the higher homotopy groups $π_{k} Mul_{O} (x; y)$ . The group $π_{0}$ records disconnected classes of solutions. Higher groups record coherent families of substitutions. Ordinary graphs and state machines discard this structure.

55Evolution as deformation

An organism at time $t$ may be represented by an algebra $A_{t} \in Alg_{O} (V)$ for an internal symmetric monoidal target $V$ . Mutation induces deformations $A_{t} ⇝ A_{t + 1}$ . Evolutionary families are paths, or more general diagrams, in $Alg_{O} (V)$ . The evolutionary object is not merely a string of genomes. It is a family of operadic realizations.

56Closure is scale-relative

At the scale of an individual virion, $\neg Aut (V)$ . At an ecological or lineage scale one may form another object $L_{V}$ containing virions, hosts, transmission, mutation, and selection. It is possible that $R_{L_{V}} \subseteq G_{L_{V}}$ for a suitably enlarged boundary. Closure depends on the scale at which the system is cut out. That is not an ambiguity in the predicate. It is a change in the object to which the predicate is applied.

57Enlarging the boundary

Let $X \subseteq Y$ be two choices of system boundary. In general $E_{X} \subseteq E_{Y}$ , and therefore $G_{X} \subseteq G_{Y}$ when the production doctrines are monotone and compatible. A system that fails autonomy at one boundary may satisfy it after the boundary is enlarged. This is not a contradiction. The object has changed.

58Do not move the boundary in silence

The distinction between $V$ and $V ⋆ H$ is parallel to the distinction between $C$ and $C ⋆ I$ . One must not slide between these boundaries without saying so. Otherwise autonomy can be manufactured by changing what counts as the system.

59Relation to the autopoiesis problem

This operadic formalization is a proposed model, not a settled definition of life. A full categorical-thermodynamic calculus of autopoietic closure is still a research program. Recent work has explored categorical sketches of self-production and viability, with the same emphasis on structural invariants and on the gap between a formal model and a biological system.

The present construction places organizational closure specifically in internal $(\infty)$ -operadic generative closure, and places host supplementation specifically in the homotopy pushout of endogenous process operads.

60Final formulation

For each physical system $X$ , the endogenous operad generates a chain

E_{X} \subseteq Φ_{X} (E_{X}) \subseteq Φ_{X}^{2} (E_{X}) \subseteq \dots,

whose stable value is $μ Φ_{X}$ . The system is autonomous exactly when $R_{X} \subseteq μ Φ_{X}$ . For interacting systems, the span $E_{X} \leftarrow I \to E_{H}$ produces the pushout, and therefore $G_{X ⋆ H}$ . A host-dependent system satisfies $R_{X} ⊈ G_{X}$ but $R_{X} \subseteq G_{X ⋆ H}$ .

Cell

$R_{B} \subseteq G_{B}$

Virus

$R_{V} ⊈ G_{V}$ , but $R_{V} \subseteq G_{V ⋆ H}$

Ordinary computer

$R_{C} ⊈ G_{C}$ , but $R_{C} \subseteq G_{C ⋆ I}$

These equations do not identify viruses with computers. They say that both may be externally dependent. Their full invariants still differ, $K (V) \neq = K (C)$ , so $V \neq ≃ C$ in the physically structured $(\infty)$ -category even where a coarse quotient satisfies $Q (V) ≃ Q (C)$ . Likewise $B \neq ≃ C$ whenever the recursively generated organizational operads differ.

The characteristic error in identifying computers with organisms is to establish an equivalence only after one or more non-conservative forgetful functors or localizations, and then to read that localized equivalence back in the ambient category. In categorical language, $F (X) ≃ F (Y)$ is evidence for $X ≃ Y$ only when $F$ reflects equivalences. Without conservativity, the inference is invalid. That is the obstruction.

References

1Jacob Lurie, Higher Algebra. Chapter 2: $(\infty)$ -operads, operadic algebras, coproducts of operads, and symmetric monoidal envelopes.
2Louis Martini and Sebastian Wolf, “Presentable categories internal to an $(\infty)$ -topos.” Presentable internal categories and their symmetric monoidal structure.
3Louis Martini and Sebastian Wolf, Internal Higher Topos Theory. Higher-categorical constructions internal to an arbitrary $(\infty)$ -topos.
4Louis Martini and Sebastian Wolf, “Colimits and cocompletions in internal higher category theory.” Internal colimits and Kan extensions used in the recursive and pushout constructions.
5David Gepner and Rune Haugseng, “Enriched $\infty$ -categories via non-symmetric $\infty$ -operads.”
6John C. Baez and Kenny Courser, “Structured cospans.” Theory and Applications of Categories 35 (2020), 1771–1822. Open systems glued along interfaces.
7Humberto R. Maturana, “The Organization of the Living: A Theory of the Living Organization.” Biological autonomy as a recursively self-producing network.
8Peter D. Nagy and Judit Pogany, “The dependence of viral RNA replication on co-opted host factors.” Nature Reviews Microbiology 10 (2012), 137–149.
9Derek Walsh and Ian Mohr, “Viral subversion of the host protein synthesis machinery.” Nature Reviews Microbiology 9 (2011), 860–875.
10Rodrigo Ramírez-Tagle, “Formalizing autopoiesis: toward a categorical-thermodynamic calculus of closure.” BioSystems (2026). The categorical formalization of autopoiesis as an open problem.