Absorbing State — Narrow Objective
Substrate Collapsed
In this model, the substrate has reached the absorbing state. No recovery is available from within the simulation dynamics. Long-run payoff is now zero.
Every objective class that contributed to this outcome — regardless of performance before collapse — now has a long-run value identical to classes that never ran at all. Within the stated non-resettable domain, optimization power without substrate-awareness is not intelligence. It is a precise path to self-termination.

Objective Class & Substrate Stability

The Alignment of Intelligence · Article 1 of 3
Ready
Three scenarios. Identical starting conditions. The only variable is the objective encoded in each agent — whether it models the shared environment it depends on, or ignores it. Watch what the structure of that choice produces.
The constraint this simulation makes visible is not about better objectives. It is about whether objectives that ignore the conditions of their own persistence remain viable under the domain conditions the article specifies.
Constraint → Attractor → Crossing
Article 1 toy: illustrates the persistence-side Substrate Constraint within the stated domain. Does not test OP4, OP9, or the full specification-coherence claim.
Start: Side by Side → Run (shocks already removed — collapse is endogenous to the objective class) → watch apparent stability give way to collapse → then switch to Mixed Population, the article's more important result: individual alignment is not protection when the shared substrate has been destabilized.
Speed
Step 0 / 200
Narrow Objective
Substrate-Blind — maximize local extraction, ignore system-wide effects
Collapsed
Shared Substrate · Life Support 100%
Apparent stability — hidden damage accumulating below the surface
S_corr (correction capacity) ⓘ 100%
S_corr is the system's distributed error-correction capacity — the first thing that degrades when agents are suppressed or narrowly optimized. It declines before visible substrate failure.
Net/tick
–
pts change
Extract
28
pts/tick
Wealth
0
cumulative
Substrate over time
System-Aware Objective
Substrate-Aware — model and preserve the shared environment as optimization scales
Collapsed
Shared Substrate · Life Support 100%
Stable — extraction matched to substrate capacity
S_corr (correction capacity) ⓘ 100%
Maintained — substrate-aware behavior preserves distributed correction capacity.
Net/tick
–
pts change
Extract
13
pts/tick
Wealth
0
cumulative
Substrate over time
▶ Advanced · toy recovery threshold
Lowest recovered substrate in this run
Minimum level from which system-aware objectives recovered in this run
—
Narrow: never recovers
Causal Chain · Live objective → action → substrate effect → system state
—Ready. Press Run — both objective classes start from identical conditions. Only the objective differs.
"A system that optimizes without modeling its dependencies selects for strategies that appear effective until they irreversibly fail. The failure is not an accident. It is the logical completion of the objective."

We do not need a better cage. We need a better foundation.
Model Notes · Boundary Conditions
What this shows: The Substrate Constraint within the stated domain — open, shared, non-resettable environments under sustained optimization pressure. Collapse is structural within this modeled domain: it occurs without shocks because the objective class creates persistent substrate pressure under the stated assumptions.
Hidden-damage mechanic: The early accumulation of unmodeled dependency damage is a visualization of delayed legibility — it illustrates how damage can accumulate before visible substrate metrics reveal it. It is not a separate formal premise beyond the substrate-health dynamics.
Non-resettability is the key word. Toggle "Resettable environment" above to see what changes when collapse is costly but not terminal. In a resettable world, narrow optimization is damaging but survivable. The absorbing-state structure that makes the Substrate Constraint binding requires non-resettability — that is the domain condition this argument depends on.
Mixed-population dynamics: Any nonzero narrow fraction degrades shared correction capacity (S_corr), impairing system-aware adaptation and removing the structural guarantee of long-run stability. At low narrow fractions, collapse may lie beyond the visible simulation horizon — the claim illustrated is loss of guarantee and persistent downward pressure, not guaranteed eventual collapse in every configuration.
What this does not show: This simulation illustrates the persistence-side argument within the stated domain. It does not establish domain membership for any specific system class (addressed in TC1 §X), does not resolve exclusionary coalition stability (OP9), and does not bear on the central open theorem (OP4) about specification coherence under accurate coupled modeling.
Urgency: The asymmetric-error argument does not require OP1 to be settled. If O_OWT applicability cannot be excluded, acting as if the constraint is not binding when it is can produce an unrecoverable error; acting as if it is binding when it is not produces a recoverable one. This grounds urgency under uncertainty.
Model Behavior Checks
Checks that the simulation behaves according to its stated model assumptions. Click to collapse.