Summary of "Out of Control: The New Biology of Machines, Social Systems, and the Economic World"

Core Idea

Kelly’s central claim is that the most important systems of the future will be vivisystems: self-organizing, self-repairing, self-improving networks in which the born and the made increasingly merge.
He argues that as systems become more life-like, they gain adaptability, resilience, and novelty, but they also become harder to centrally control.
The book’s big inversion is that life, not machine-like clockwork, is the better model for complex future technology, organizations, and even planetary systems.

Distributed intelligence is a recurring theme: bee hives, flocks, markets, brains, and networks coordinate through local rules, sparse communication, and feedback rather than top-down command.
Kelly treats the hive as a superorganism and audience-controlled demos, flocking algorithms, and swarm robotics as proof that collective order can emerge from many simple agents.
His key concept is emergence: higher-level properties appear that cannot be inferred from the parts alone, so the system must be run to be understood.
He links memory and perception to distributed reconstruction rather than stored files, using clinical amnesia cases and sparse distributed memory research as evidence.
The book repeatedly contrasts clockware with swarmware: use the former when supreme control matters, the latter when adaptation, robustness, and evolution matter more.
Network logic is central: the Net replaces the Atom as the era’s icon because it captures interdependence, hubs, flows, and no single center.
But networks are not pure anti-hierarchy; Kelly emphasizes nested hierarchies and layered control, with fast local actions feeding slower higher-level structures.

Kelly’s most ambitious move is to say that nature and technology are coevolving into one another: organisms are engineered by humans, while machines borrow biological traits like replication, learning, self-governance, and repair.
He calls gardens, economies, corporations, robots, circuits, and ecologies alike vivisystems because they all exhibit some degree of life-like organization.
Coevolution is a major explanatory engine: monarchs and milkweed, chameleon and mirror, ants and acacia, and even markets and strategies all adapt in response to each other.
He treats Earth itself as a vivisystem/Gaia: the atmosphere, rocks, and living matter form a coevolving, self-regulating whole in persistent disequilibrium.
The Gaian point is not that the planet is conscious, but that life and geology jointly regulate conditions in ways that look like a distributed thermostat.
Kelly’s evolutionary stance is that life is a form of learning: not magical foresight, but iterative adaptation, feedback, and selection across interacting systems.
He extends this to culture and game theory, where repeated interaction makes openness, reputation, and Tit-for-Tat-like cooperation more effective than secrecy in many nonzero-sum settings.

Chapter by chapter, Kelly argues that complex systems generally cannot be fully designed in one shot; they must be assembled incrementally, often with temporary scaffolds and indirect control.
The ecological restoration examples show that successful ecosystems depend on the right sequence, chaperone species, disturbance regimes, and hidden stowaways, not just the right inventory of species.
The reef, prairie, Bermuda island, and Biosphere 2 examples all show that assembly order matters and that some species or supports are only needed during startup.
His closed-world experiments suggest that a system can be materially sealed yet remain viable if it can close the bio-elemental loops and preserve enough turbulence to prevent stagnation.
Biosphere 2 becomes the emblem of this idea: a large, closed, energetically open ark whose fate depends on whether human engineering can eventually hand control over to biology.
In robotics, this logic appears as Rodney Brooks’s subsumption architecture and his vision of fast, cheap, and out of control mobots that act embodied in the real world rather than relying on a central model.
Kelly insists that embodiment matters: intelligence without a body is misleading, because sensory deprivation and disembodied computation drift into hallucination and fantasy.

Kelly repeatedly recasts evolution as search through a vast space: Borges’s library, Karl Sims’s evolving images, Dawkins’s biomorphs, Latham’s forms, and Koza’s genetic programming all model discovery as guided variation.
A central lesson of artificial evolution is that search can become creativity when the space is large enough and the selection process is good enough.
He favors variable-length, open-ended systems over fixed-gene systems because open genomes and recursive primitives allow richer evolutionary possibilities.
The artificial life movement, from Game of Life to Tierra, argues that life is a process, not a substance, and that new life-like systems can be built to study life itself.
Kelly is especially interested in hyperlife: a broad category of strong, cohesive vivisystems that includes rain forests, networks, simulations, and machines.
In his view, evolution is not just changing but becoming more evolvable: more complex, diverse, codependent, specialized, and capable of generating novelty.
He also argues that organisms and organizations can learn before genes catch up, so flexibility, culture, and adaptation can steer long-term evolutionary outcomes.
The endpoint is not a final perfected form but perpetual novelty: systems that keep generating new possibilities, including machines that evolve their own architectures.

The book’s deepest argument is that the future belongs to systems that are self-organizing, embodied, and coevolutionary, not centrally optimized machines.
Kelly treats control as a limited and often counterproductive ideal in complex domains; survival comes from designing conditions under which systems can grow, adapt, and surprise us.
His broadest synthesis is that biology is the prototype for postindustrial design, and the ultimate technology may be less like a machine and more like a living ecology.
The book ends by making evolution itself the model: not a ladder of progress, but an open-ended process whose real achievement is the creation of possibility.