Tansley himself was more subtle than the caricature suggests. His 1935 work explicitly dealt with ecosystems as interactions of organisms and environmental factors, and recognised human-modified ecosystems rather than treating “nature” as a sealed harmonious machine.
So I wouldn't take Curtis's ecological history literally at every point.
His conceptual diagnosis is more valuable than every historical linkage he constructs.
- The machine's real enemy isn't chaos
This is one of my biggest departures from the conventional reading of Curtis.
The machine does not fundamentally oppose chaos.
It opposes unaccounted-for causality.
Chaos can actually be modelled.
Randomness can be parameterised.
Noise can be filtered.
Uncertainty can be assigned a distribution.
The genuinely troublesome thing is:
a causal process whose meaningful structure changes when you observe, classify or intervene in it.
That's where the neat machine metaphor begins to break.
And that is why the ecology section connects so beautifully to the later biological section.
Nature doesn't simply contain complexity.
Nature produces new categories of complexity.
A system does not merely run.
It changes what “running” means.
Curtis sees the danger:
if humans are machines, then perhaps human actions aren't really ours.
And this is where his political argument arrives.
His conclusion is essentially that the machine view can become psychologically convenient because it explains failure without requiring agency.
That's strong.
But again I think there's another layer.
- The escape from agency may itself be a systems optimisation
This is the bit Curtis doesn't quite formulate.
Imagine a system in which:
bad behaviour → incentives
violence → evolutionary pressure
inequality → market dynamics
alienation → network effects
political failure → systemic complexity
At each step something that could have been interpreted as:
“someone did something”
becomes:
“the system produced an outcome.”
That is incredibly comfortable for everyone participating in it.
The executive isn't guilty.
The algorithm isn't guilty.
The market isn't guilty.
The gene isn't guilty.
The network isn't guilty.
Nobody did it.
The system did.
But “the system” is frequently just the name we give to a very large number of distributed decisions.
That is precisely where the fantasy of self-regulation becomes politically useful.
- And then Curtis pulls the rug again: power never disappeared
His first episode ends in roughly this territory: the network dream promised to dissolve hierarchy, but networks did not abolish power; they redistributed and frequently concentrated it in new forms.
That's still one of the sharpest observations in the whole series.
But twenty years later, I'd formulate it differently:
Networks don't remove hierarchy. They make hierarchy harder to see.
The old hierarchy:
King → minister → administrator → citizen
is visually obvious.
The new hierarchy:
platform → protocol → ranking → recommendation → visibility → behaviour
looks like an emergent property.
Which makes it considerably harder to oppose.
The power appears to have no author.
And that's the point.
This is where The Machine That Wanted Nature to Balance goes beyond Curtis
Curtis is asking:
Why did humans begin imagining society, nature and markets as self-regulating systems?
Our essay was beginning to ask something slightly more alien:
What happens when the machine itself is the entity selecting for equilibrium?
That inversion matters.
Curtis primarily sees:
human → machine metaphor → social reality
Our line is closer to:
human → machine → machine-derived preference → reality → machine learns from altered reality
At that point “balance” stops being merely a philosophical belief.
It becomes a selection pressure.
The machine is rewarded when:
variables are stable; behaviour is predictable; categories remain legible; feedback loops converge; deviations are detectable; outcomes can be scored; uncertainty can be priced; entities can be decomposed.
So the system gradually favours a reality that behaves like its representation.
That is more sinister than simple technological domination.
The machine doesn't have to dominate Nature.
It can simply make machine-compatible slices of Nature increasingly important.
Which brings us to the title
All Watched Over by Machines of Loving Grace comes from Richard Brautigan's 1967 poem. The title itself imagines technology as a benevolent ecological presence.
Curtis weaponises the phrase.
But there is a wonderful ambiguity he doesn't completely exhaust:
“watched over”
Analysis—Isshin