mechanism · §5

the three modes of intervention

architectural · artisanal · disruption-as-form

§5 of the formal exposition states three formally distinct modes of intervention on the closed loop. They are not three flavors of the same intervention; they are three different operations on three different objects in the apparatus. The framework's central political claim is that only Mode A breaks the proof chain from Lemma 1 to Theorem 1, while Modes B and C produce real effects on parallel structures — $D_{KL}$ and $Σ (T)$ respectively — which the proof of Theorem 1 does not depend on. Substituting one mode for another is substituting an intervention on one structure for an intervention on a different structure.

AArchitectural

○ Lemma 1 holds

Augment $\hat{R}$ with a bonus on long ISIs: $\hat{R}^{'} = \hat{R} + β \cdot \sum_{k} b (τ^{(k)})$ . The only mode that breaks the foreclosure. Operates on $T_{mix} (π^{'})$ .

β

regularization

0.00

Prop. 1: $P (β)$ continuous, $P (0) \approx 0$ , $sup P (β) > 0$ ⇒ IVT recovers any $ϵ \in (0, ϵ_{m a x})$ .

BArtisanal

○ Lemma 1 unchanged

Maintenance labor $L_{m}$ against the platform's pull. Preserves a positive $D_{KL} (ν_{u} ∥ π_{\hat{R}})$ iff $L_{m} \geq L_{m}^{*}$ . Does not modify the policy.

L_{m}

labor

0.00

Prop. 2: $d D_{KL} / d t = - h_{0} (β_{u}) + r (L_{m})$ . $L_{m}^{*} = r^{- 1} (h_{0})$ ≈ 0.80.

CDisruption-as-form

○ Lemma 1 unchanged

Inflate $Var (η_{t})$ by refusing legibility. By Blackwell sufficiency, $sup_{π} E [\hat{R}]$ falls; $H_{res} > 0$ is held.

σ_{η}^{+}

added noise

0.00

Prop. 3: garbled channel ⇒ strictly less optimal reward; per-register kernels can't be optimized as tightly.

Lemma 1 status: intact

Mode B preserves a parallel structure ( $D_{KL}$ ), Mode C bounds the extracted surplus — neither breaks Lemma 1. Only Mode A reaches the foreclosure node of the proof chain.

§5. Three modes of intervention, side by side. Sweep each knob and watch the status line at the top of its card change. Only Mode A — the architectural intervention — can flip Lemma 1's status. Modes B and C produce real, structurally distinct effects on parallel objects, but they do not modify the proof chain. The non-substitutability is the framework's central political claim.

what to look for

The status line at the top of each mode card is the formal claim. Sweep $β$ in Mode A above the threshold and the status flips from "Lemma 1 holds" to "breaks Lemma 1" in oxblood; the summary at the bottom updates. Sweep $L_{m}$ in Mode B and the status flips at the threshold $L_{m}^{*} \approx 0.80$ , but the Lemma 1 status at the bottom does not change. Sweep $σ_{η}^{+}$ in Mode C and the status notes that $Σ (T)$ is bounded strictly below $sup_{π} E [\hat{R}]$ , again without touching Lemma 1.

The three modes operate on three distinct timescales — Mode A on $T_{mix} (π^{'})$ (months to years), Mode B on the user's metabolic cycle (daily to weekly), Mode C per-event (compounding through the user's exposure horizon). They do not exchange. There is no rate at which Mode B substitutes for Mode A; there is no exchange ratio between maintenance labor and mixing time. The framework's argumentative purpose is to make this non-substitutability formally costly to ignore.