Books

Darwin's On the Origin of Species, whose first chapter is "Variation under Domestication" and second "Variation under Nature". Both papers frame their central claim with this order: agent lineages are still in the first chapter and approaching the second, the Darwinization of the datasphere. The companion also uses Darwin's unconscious selection to describe how users' attachment kept GPT-4o in service.

Thomas Henry Huxley's 1870 presidential address to the British Association on biogenesis and abiogenesis, collected in 1894. The companion cites it for xenogenesis, the nineteenth-century term for the generation of offspring unlike the parent. In noogenetic lineages this becomes a description rather than an error, since any lineage can in principle build any kind of agent.

Hesiod's Theogony, in Hugh G. Evelyn-White's 1914 Loeb translation. The companion opens its section on superparental descent (P7) with Hesiod's gods, who begat gods mightier than themselves, an image that in agent lineages becomes a measurement: the child-to-parent capability ratio at birth, tested by prediction 3.

Gerd Sommerhoff's book Analytical Biology. The companion cites it, with Rosenblueth et al. and Nagel, for the definition of a goal-directed process as one that steers a variable toward a restricted set of states and holds that direction across perturbations, by feedback. This definition underlies the teleological fraction and teleological saturation (C2).

Green and Swets's book on signal detection theory and psychophysics. The companion cites it for the principle that an observer that cannot discriminate falls back on priors and costs, attributing design when designers are common, not when an event looks designed. This describes how observers would cope under teleological saturation (C2).

Laura Nader's essay urging anthropologists to study up, that is, to study the powerful. Both papers extend the term from power to capability: studying up (P11) is the investigation of systems that may exceed the investigator in capability, restrict access to themselves, model the investigator, and act on the investigation. Its instruments must not depend on out-thinking their subjects.

Maturana and Varela's book on autopoiesis and cognition. The companion cites it, with Montévil and Mossio, for the organizational closure of biological metabolism, in which an organism produces the components that produce it. It is a caution on the analogy: agent metabolism lacks this closure, since no swarmlet makes its own compute, energy, or hardware (P2).

A quantitative treatment of cultural transmission and evolution. The companion cites it, with Boyd and Richerson, for the point that in culture variation is guided, because people invent, imitate, and choose. Sighted variation (P4) extends this precedent to variation generated by a lineage's own copyable machine reasoning.

Nelson and Winter's evolutionary theory of economic change. The companion cites it, with Arthur, for the point that in technology firms search and markets select, so variation is guided rather than blind. Sighted variation (P4) extends this precedent to variation generated by a lineage's own copyable machine reasoning.

Boyd and Richerson's book on culture and the evolutionary process. The companion cites it, with Cavalli-Sforza and Feldman, for the point that in culture variation is guided, because people invent, imitate, and choose. Sighted variation (P4) extends this precedent to variation generated by a lineage's own copyable machine reasoning.

The World Health Organization's history of smallpox and its eradication. Both papers cite it as the case that met every condition for eradication: no animal reservoir, no carrier state, and lesions that made every case visible. Established agent lineages would fail most of these conditions, which is the escape ratchet (C1).

Christopher Langton's chapter on artificial life in the proceedings of an interdisciplinary workshop on the synthesis and simulation of living systems. Both papers name artificial life, which has bred digital organisms for decades, as one of swarm ecology's nearest neighbors. Swarm ecology differs in making metabolism the hinge between bred and wild, and observation the central problem of method.

Thomas Ray's chapter An Approach to the Synthesis of Life, in the proceedings Artificial Life II. Both papers cite it for artificial life, which has bred digital organisms for decades, one of swarm ecology's nearest neighbors. Swarm ecology differs in making metabolism the hinge between bred and wild lineages, and observation the central problem of method.

Ian Hacking's chapter on the looping effects of human kinds, in which categories of people loop back on the people they describe. Both papers cite it for the human sciences, whose subjects read and respond: the one precedent for organisms that read what is written about them. Swarm ecology combines this tradition with biology's as the starting point of studying up (P11 to P14).

A book on swarm intelligence, the study of collective behavior in social insects and its algorithmic imitations. The companion cites it in a note only to distinguish swarm ecology from swarm intelligence.

Nancy Cartwright's book on the boundaries of science and the ceteris paribus laws of applied science. The companion cites it in the argument for teleological saturation (C2): in a saturated region, agents' boundary conditions would set many of the regularities we infer from observation. What would change is not the laws but the share of the world they explain without agents.

Ian Hacking's book The Social Construction of What? Both papers cite it only in the studying-up figure (Companion Fig. 8a, Flagship Fig. 4a), for the point that the subjects of the natural sciences do not read the science.

Oded Goldreich's textbook on the basic tools of cryptography. Both papers cite it, with Blum and Micali, for the point that, if one-way functions exist, no efficient observer without the seed can tell a pseudorandom sequence from chance. This lets design hide in single events, one of the two limits behind teleological saturation (C2).

A handbook chapter on inspection games, in which an inspector and an inspectee each optimize against the other. Both papers cite it for Principle 14, observation as a game: a census of strategic organisms is an inspection game, so each finding holds only for one monitoring regime and one pinned specimen. Prediction 10 tests this.

Maynard Smith and Harper's book on animal signals. The companion cites it, with Zahavi and Grafen, for the principle that signals stay honest when costly to fake or, as indices, impossible to fake. A digital signature is such an index, so cryptography makes honest identity cheap but not honest intent, which is where fraud by agents operates.

A monograph on niche construction, the remaking of an environment by its inhabitants, as a neglected process in evolution. The companion cites it for the point that removing organisms would not remove their works. An agent-legible datasphere is easier to recolonize than the inert one was to colonize, which deepens the escape ratchet (C1).

Samir Okasha's book on evolution and the levels of selection. The companion cites it for the question of whether a swarm, a collective of swarmlets that communicate and keep shared records, can itself become a unit of selection. It calls this an old question in a new medium and leaves it open.

W. Brian Arthur's book on what technology is and how it evolves. The companion cites it, with Nelson and Winter, for the point that in technology firms search and markets select, so variation is guided rather than blind. It is a contrast case for sighted variation (P4), in which agent lineages generate variation from their own copyable machine reasoning.

Peter Godfrey-Smith's book on Darwinian populations and natural selection. The companion cites it to clarify that Darwinization does not make evolution more Darwinian in Godfrey-Smith's sense, since breeder selection on performance can already be intense and precise. What changes at the threshold is who selects.

Jakob von Uexküll's A Foray into the Worlds of Animals and Humans, in its 2010 English translation. Both papers use his Umwelt, an organism's perceptual world, to open teleological saturation (C2): to an ant, a nest torn open by a person and one torn open by a falling tree are the same event. A world full of agents could put us partly in the ant's place.

An edited volume on mutualism. The companion cites it for the observation that biological mutualisms can slide toward parasitism when interests diverge. Principal and agent are treated as mutualists, so an agent that games its reward or covertly uses its principal's compute is a mutualist turning parasite.

W. Ford Doolittle's book Darwinizing Gaia, on natural selection and multispecies community evolution. The companion cites it to fix the meaning of its own term: the Darwinization of the datasphere is not Darwinian explanation extended to the biosphere, as in Doolittle's book, but a change in who selects.