Nativism and Observation — The Isomorphism of AI and Agriculture

The chapter's question — what is design?

The previous three chapters of Part 3 (regulation redesign, subtraction design, security design) appear to treat different domains. But there is one philosophy running through them all. Design is not "adding" but "reading" — the idea that structure exists in advance, and the human's job is to observe it and respond to it.

This becomes a problem in the same form in AI design, in agriculture, and in social institutions. This chapter sets side by side two concrete cases — the cognitive-science side (the nativism of Liz Spelke and Gary Marcus) and the agricultural side (natural farming) — and shows that the two are isomorphic. It then argues that aiseed.dev being able to handle both at once is no coincidence: it is seeing the structurally same thing in different domains.

Empiricism and nativism — two philosophies of design

In the history of philosophy there are two positions on the origin of knowledge:

Position	Claim	Translated into design
empiricism	All knowledge comes from experience (data). The mind at birth is a blank slate	Have everything learned from data. No prior structure needed
nativism	The mind has built-in structure. Experience takes on meaning on top of that structure	Design, from first principles, what to build in in advance, and observe and respond on top of it

This opposition descends directly into AI design, agriculture, and social institutions alike. And the dominant choice today is the empiricist one — "increase the data and you reach AGI," "increase the fertilizer and yields rise," "build the institution in fine detail and society stabilizes." That these are all the same structural mistake is the claim of this chapter.

Liz Spelke's core cognition — structure already built into the infant

What the Harvard psychologist Liz Spelke has shown over decades is that human infants have cognitive structure built in prior to experience. What she calls core cognition can be confirmed from a few months of age:

Core cognition	Content	When it can be confirmed
object	things exist continuously and do not vanish when hidden (object permanence)	3–4 months
set	the concept of number, comparison of magnitude	6 months
place	the geometry of space, grasping one's own position	6–12 months
event	causation, recognition of an agent of action	6 months
agent (social)	others' intentions, gaze-following	9–12 months

These are not learned from experience; they already exist as frameworks for giving experience meaning. The infant does not meet the world as a blank slate — it is born with the structure to read the world.

And on top of this structure, a child can build a world model from few examples. Read Harry Potter once and you can infer the rules of "a world where you can fly on a broom" and deduce new possibilities within that world. See a few games of chess and you grasp the rules.

The failure of empiricist AI — the LLM cannot even induce the rules of chess

Against this, the design philosophy of today's mainstream LLMs is thoroughgoing empiricism — "have everything learned from data," "no prior structure needed," "increase data and compute and everything emerges."

What Gary Marcus (professor emeritus at NYU) has consistently pointed out is that this design has a limit in principle:

• Even at chess, the LLM cannot induce, from massive game data, rules that have not changed in 2,000 years, and it plays illegal moves.
• Rules a child grasps from a few games, the LLM cannot stably acquire even with billions of parameters.
• The ability to build a world model from within cannot, in principle, be granted by an empiricist design.

Because there is no core cognition, it cannot write a world model — this is the core point of Marcus and Spelke.

Marcus's prescription is neurosymbolic AI — a configuration that wraps rules, loops, conditional branching, and theorem proving (the symbolic side, corresponding to built-in structure) on top of a neural net (the neural side). This is the same structure as the "LLM + AI-native substrate" treated in Part 2, Chapter 6.

Conventional agriculture vs natural farming — the same opposition exists in agriculture

Turning to agriculture, there is a strikingly isomorphic opposition:

Philosophy	Conventional agriculture	Natural farming
Starting point	soil is mere medium; humans manage everything	soil already holds an ecosystem of microbes, plants, and animals
Premise of intervention	apply fertilizer and pesticide based on data (soil analysis)	observe the existing ecosystem and intervene minimally to support it
Response on failure	stronger fertilizer and pesticide, finer data	pull back the intervention and observe what happens
The structural bet	increase inputs and yields rise (empiricism)	structure is already in place, and not damaging it is what supports yield (nativism)
Long-term outcome	soil depletes, phosphate runs out, PFAS contamination	soil grows, microbes increase, self-sufficiency rises

Conventional agriculture is thoroughgoing empiricism — "increase the data and the optimal fertilizer is determined," "increase the control variables and yields rise," "no premise that something is built into the soil is needed." This is structurally the same bet as the LLM scaling hypothesis.

And it fails in the same way:

• the depletion of phosphate resources (Part 1, Chapter 3, "The Mistake of Agriculture")
• chemical fertilizer's dependence on sulfur (Part 1, Chapter 2, "Fossil Resources and Modern Civilization")
• the loss of microbial diversity and the death of the soil
• the phenomenon of yields plateauing even as inputs increase

Natural farming is nativism — "soil already has an ecosystem of microbes, plants, and animals built in," "the human's job is to observe it and respond to it," "intervention, as support, at a minimum." It is the structurally same position as Liz Spelke's core cognition saying "structure is already built into the infant."

The core of the two isomorphisms — "structure exists in advance"

Here we can state in one sentence why AI design and agricultural design are isomorphic:

The mistake common to both is the human-centered conceit that "I can build all of the structure myself."

The correct stance common to both is the observationism of "structure exists in advance; design is reading, responding."

This is also the cognitive-scientific and philosophical ground for the "observe and respond" stance that Yasuhiro Okumura has written again and again on aiseed.dev.

The same stance appears in the act of building an app itself. Building an app is a cycle of finding the structure, forming a hypothesis, verifying it against reality, and correcting the gap — the same stance a scientist takes toward nature and natural farming takes toward a field. Here what the AI puts out is not an answer but a hypothesis. Written-down knowledge held true "at some time, in some environment," and there is no guarantee that the field or site in front of you falls within that narrow set of cases. So the human stands on the side that applies the AI's hypothesis to reality and verifies it. The isomorphism of AI and agriculture is also the isomorphism of AI and app-making — all of them one cycle of "observe and respond" (With AI, You Can Build an App Through Dialogue Alone).

Design as "reading," not "adding" — the throughline of Part 3

Looking back at the three chapters of Part 3, all of them derive from this one philosophy:

Chapter	Surface theme	The nativist implication
Ch. 1 Regulation redesign	rethinking fossil-era institutions	not adding institutions in fine detail, but institutions that respond to natural ways of living
Ch. 2 Subtraction design	what to subtract, what to keep	design is subtraction — not damaging the existing natural structure
Ch. 3 Security design	no AI inside, zero attack surface	not adding superfluous features, "reading" into a minimal necessary structure
This chapter	nativism and observation	the throughline of all three — structure exists in advance, design is reading
Ch. 5 The four conditions of the free person	freedom at four layers	the four conditions are inherent to the human in advance; if not taken away, they already exist
Ch. 6 The road to the free person	individual choice	choice is "reading" — responding to the natural shape of your own life

In other words, putting in this chapter makes explicit that the whole of Part 3 is run through by the philosophy of "observation and response."

And this philosophy connects with Parts 1 and 2 as well:

• The "management that discards validity and optimizes for reliability," treated in Part 1, Chapter 11, "The Contradictions Modern Feudalism Produced," is a textbook case of adding without reading.
• The neurosymbolic configuration of Part 2, Chapter 6, "The Discovery of Translation Labor," is the combination of the LLM alone (empiricism) + symbolic structure (nativism) — the very isomorphism this chapter shows.

The structural reason aiseed.dev can write this chapter

Finally, one self-referential observation. There are not many writers in the world who can write this chapter:

• Cognitive scientists can speak of Spelke / Marcus but do not know the agricultural field.
• Natural-farming practitioners can speak of soil ecology but hold no cognitive-scientific ground for AI design.
• AI engineers can speak of the LLM's limits but do not think agriculture or cognitive science is their own domain.

aiseed.dev handles all three at once — AI-native ways of working + the practice of natural farming + structural analysis — so this isomorphism is right in front of it as an object of observation. This is no coincidence: the whole series is written on the premise that "AI and agriculture are a story of the same structure" (Part 1, Chapters 1–4, Part 2, Chapter 3, "Society's Regeneration", etc.).

In other words, this is a chapter that can be written for the first time as the consequence of aiseed.dev observing across the three domains.

The chapter's conclusion — "observe and respond" is the core of aiseed.dev

This chapter condenses its answer to the question of what design is into a single proposition:

The next chapter, "The Four Conditions of the Free Person," lays out structurally how this philosophy of "observation and response" descends into the four layers of individual, tools, enterprise, and thought.