Total Structure v3: Law X — FALSIFICATION

The Eight Surfaces

S1 — The moral floor. Produce one full-scope case where tolerated remediable subjugation genuinely increases efficiency after enforcement cost, externality, recurrence, suppressed capability, downstream instability, and maintenance burden are counted. Full scope is bounded, so the falsifier is difficult but not rigged. Success collapses the identity claim, the triple optimum, and everything downstream of A₄ — which is everything.

S2 — The convergence claim. Produce a genuine full-scope value conflict — ethics against efficiency, truth against utility — that survives complete accounting without dissolving into incomplete scope, false boundary, or omitted cost. Success collapses A₃ and reduces the structure to one more balancing act.

S3 — The decay clock. Show that predation tolerance is not a leading indicator of systemic decay — that societies with rising tolerance for remediable harm against the unremedied do not subsequently exhibit the decay signature, or that the correlation runs the other way. The clock is stated as measurable; measure it.

S4 — The machine plane. Show that unscaffolded stochastic inference consistently produces higher task-adjusted logical density than deterministic scaffolding on audit-dependent tasks, after coordination, verification, latency, and human-review costs are counted.

S5 — Amortization. Show that proof artifacts fail to amortize in practice: verification exceeding regeneration, similarity classes too rare, freshness windows too short, artifacts non-transferable across actors without loss of validity. Success weakens the deterministic-era claim to "marginal improvement on some tasks." Status note: the running build demonstrates single-operator reuse; the cross-actor market claim remains open on this surface — the structure says so itself in Law VI.

S6 — The command plane. Show that LLM-as-OS cannot operate at scale — routing overhead exceeding task-adjusted gain, unavoidable control-plane capture, meta-decisions that cannot be made glass, structural isolation unmaintainable under realistic adversarial conditions. Success collapses the machine implementation to scaffolds-per-task.

S7 — The object grammar. New in v3.0. Show that the universal loop fails as a load-bearing pattern: receipts forgeable or ledgers tamperable at realistic cost; repair lineage unmaintainable at scale (fixes detaching from failures faster than discipline can reattach them); the one-door pattern becoming the single point of capture it claims to prevent; or the zero-context rule unachievable — self-description collapsing under real complexity into the very interpretive priesthood it was built to delete. Success reduces Law VI from doctrine to one system's housekeeping.

S8 — The recursion. New in v3.0. Show that self-review loops degrade rather than improve documents under sustained operation: reviewers Goodharting the clarity scores; machine revisions drifting semantic content while polishing surface; append-only version chains accumulating noise faster than signal; or the review loop itself being captured through reviewer selection. Success collapses A₁₂'s constructive half and reduces Law IX from living protocol to versioning ritual.

Higher surfaces can fail without collapsing lower ones. S1 demonstrated collapses everything. An attack that engages none of the eight is not an attack on the operational core.

The Attack Protocol

A valid attack does six things:

1. Engage the strongest version. State the claim back in its strongest form, qualifiers intact, before attacking. Attacking a weakened restatement is not engagement.
2. Name the surface. State which of S1–S8 the attack engages, and what survives if it succeeds.
3. Name the exact claim. Quote the line. Diffuse criticism of the general orientation is not an attack.
4. Classify the attack. Definition, logic, empirical, scope, category-error, implementation, prior-art, or falsifiability. Multiple types may apply; name them.
5. Show full-scope accounting. Where empirical, show the costs: enforcement, externality, recurrence, suppressed capability, downstream instability, maintenance burden, audit debt. A counterexample that excludes a known cost is a scope error, not a counterexample.
6. Propose the minimum patch. If the attack succeeds, what is the smallest revision that lets the structure survive? An attack without a minimum patch is a demolition request, not engagement.

A surviving attack's minimum patch enters Law IX as a Class R amendment. This is the full loop: the attack protocol is not the document's defense — it is the document's intake.

---

---

Corpus map

• Canonical shelf: Total Structure root