Monitoring vs. Engaging

In the AI Story

Monitoring is supervisory. The monitor surveys the process without participating in it. Her attention is distributed across the operation, scanning for anomalies, evaluating outcomes, ready to intervene when deviation occurs. Her stance is evaluative — asking at every moment whether output meets standard, whether process is on track, whether what she observes requires correction. The monitor's contribution is real: quality assurance, error detection, the maintenance of coherence across complex systems. But the monitor is not inside the work. She is outside it, assessing from a position that prevents the immersion that generates understanding.

Engaging is participatory. The engaged person is inside the work. Her hands are on the material. Her attention is concentrated at the point of contact between intention and realization. She is making decisions in real time, not evaluating decisions already made. She experiences the resistance of the material, the friction of the medium, the specific quality of struggle that arises when what she wants to achieve and what the material permits do not align. Before AI, the builder was inside the work. She wrote the code, crafted the prose, designed the interface. Each encounter with resistance deposited a layer of understanding that accumulated over years into the embodied expertise distinguishing master from novice — the sense that something is wrong before you can articulate what, the architectural intuition that cannot be extracted from documentation because it was never in the documentation. It was in the hands.

AI shifts the builder from inside to above. Instead of writing code, she monitors the AI's code. Instead of crafting prose, she evaluates the AI's prose. The shift is gradual and individually rational — the AI writes faster and often better, the monitoring posture produces more output per unit time. Over time the balance tips. The engagement mode — the mode in which understanding is built through friction with resistant material — is used less frequently. The monitoring mode becomes default. And the default, once installed, is self-reinforcing: the more the builder monitors, the less practiced her engagement becomes, the less practiced engagement feels less productive than monitoring, and the less she does it.

In The Orange Pill, the senior engineer in Trivandrum illustrates this dynamic precisely. He recognizes that the twenty percent of his work the AI cannot assume — judgment, architectural instinct, taste — is the part that matters. The recognition is correct. But Stone's framework identifies the hidden dependency: that twenty percent was formed during the eighty percent the AI has now assumed. The judgment was not a separate faculty that happened to coexist with implementation skill. It was the residue of implementation — the accumulated deposit of thousands of encounters with resistant material, each adding a thin layer of understanding to the foundation on which judgment rests. When implementation is delegated to the machine, the residue stops accumulating. The existing judgment continues to function. But the judgment is not being replenished, and judgment, like any cognitive capacity, requires exercise. The exercise that built it was the engagement the AI has replaced.

Origin

Stone developed the distinction through her decades of observing attentional postures in technology environments, recognizing that the subjective experience of intense engagement could correspond to two structurally different cognitive states — one that produced understanding and one that produced output without understanding. The framework draws on Richard Sennett's The Craftsman (2008) and his distinction between workmanship of risk and workmanship of certainty.

The application to AI follows directly: AI shifts the builder from the workmanship of risk (where outcome depends on the maker's skill and material contingency) to the workmanship of certainty (where the process determines the outcome). The cognitive and emotional engagement that risk produces — investment, presence, attention — is transferred from builder to machine, and with it the developmental pathway through which judgment forms.

Key Ideas

Watcher and maker are categorically different. The watcher may understand the process intellectually but has not undergone the experience that produces embodied expertise.

Monitoring is supervisory; engaging is participatory. The two postures are not points on a spectrum but structurally different relationships to the work.

Judgment is residue of engagement. The capacity to evaluate is built by years of doing, not by years of evaluating — the very twenty percent the AI cannot replace was formed by the eighty percent it now assumes.

Self-reinforcing default. Each shift toward monitoring makes engagement less practiced, less productive-feeling, and less likely — installing supervision as the only mode the builder maintains.

Capital consumed without investment. Existing judgment continues to function but is not being replenished; the foundation on which AI-augmented work depends is being drawn down without deposit.

Debates & Critiques

Some argue that judgment can be cultivated through evaluative practice itself — that the AI-era builder develops new forms of expertise through skilled monitoring of machine output. Stone's framework allows this possibility but insists that evaluation-only expertise differs in kind from expertise built through engagement, and that the difference shows up in the limit cases where novel problems require the embodied intuition only sustained engagement deposits.