Autopoiesis as Maturana strictly defined it exists in the molecular domain, but the organizational logic — a system whose product is itself — applies to the nervous system as well. Cognitive self-production is the continuous generation of the knower through the very activity of knowing. Each episode of effective action modifies the organism's neural structure, attentional habits, and embodied response patterns in ways that persist and shape all subsequent interactions. The engineer debugging a system does not only solve the problem — she produces herself as the kind of engineer capable of recognizing and solving that class of problem the next time. The layers of understanding Segal describes depositing through years of engagement are the geological record of this self-production. It is not that skills accumulate; it is that the self that performs the skills is made through the performing.
The concept is Maturana's framework applied one level up from its strictest formulation. In 2002 Maturana reiterated that autopoiesis proper exists only in the molecular domain — extending it metaphorically to social systems or corporations misleads. But cognitive self-production at the level of the nervous system is not metaphorical. The nervous system is a biological system whose own operations modify its own structure, producing the continued existence of the knower through the knower's activity. This is autopoiesis operating at a higher level of biological organization, not a metaphorical extension.
The concept becomes crucial for understanding what is at stake in AI-assisted work. The engineer in Trivandrum had produced herself over eight years through recurrent effective action in backend systems. Her nervous system bore the structural modifications that constituted her knowing. When Claude Code arrived, her output expanded dramatically — she built interfaces she could not have built before. But the mechanism of self-production shifted. Where she had produced herself through the struggle of implementation, she now produced artifacts through conversation with a system handling implementation on her behalf.
Segal observed this directly: a senior engineer spent months making architectural decisions with less confidence than he used to, and could not explain why. The explanation in Maturana's framework: the rare, formative moments within implementation hours — when something unexpected forced genuine learning — had been absorbed by the machine. The engineer's layers of self-production had stopped being deposited in the specific domain where his confidence was built. The artifacts kept coming. The cognitive self-production that had constituted him as that kind of knower had quietly diminished.
The cost is not loss of skill but loss of the activity through which the knower is made. The distinction matters: a builder can remain capable of an action while losing the capacity to become the kind of being for whom that action is a natural extension of her self-making. She may perform the action by prompting a machine to perform it, producing output indistinguishable from the pre-AI case, while undergoing no structural modification. Her capability is borrowed from the system; her self-production is interrupted.
The concept is a direct implication of Maturana's framework but the specific phrase 'cognitive self-production' is a contemporary extension, drawing on Evan Thompson's 'Mind in Life' (2007) and the broader enactive tradition that has developed Maturanian biology into a theory of mind. The phrase captures what Maturana himself sometimes called 'the observer in the languaging' or 'the self arising in the act' — the continuous generation of the knower through knowing.
Edo Segal's application of the framework in this volume — the insight that the dashboard showed his engineer thriving while her body said otherwise — identified the specific mechanism without initially having the vocabulary. Maturana's framework provided the language: what was happening was the interruption of cognitive self-production by the delegation of the activities that produced it.
The self is made through knowing. Not that the self has knowledge but that the self is the ongoing production of the knower through effective engagement.
Each act deposits the knower. Individual episodes of effective action are not instances of using a stable knower; they are events in which the knower is continuously produced.
Delegation can interrupt production. When the activity through which the knower is made is performed by an allopoietic system on the builder's behalf, artifacts accumulate but self-production can quietly stop.
The missing metric. Productivity dashboards measure output, not self-production. The engineer whose output rises while her self-production diminishes looks, by every visible measure, like she is thriving — and the contradiction is invisible until the consequences surface.
Whether cognitive self-production can be maintained in a coupling where implementation is delegated depends on whether the builder finds new domains of effective action sufficient to sustain her autopoiesis. The ascending friction thesis predicts she can. The Berkeley study suggests she often does not — that freed time fills with more delegation at the same level rather than deeper engagement at a higher level.