Technical Bridge
The one claim
Bilateral validity does not guarantee global coherence. The research program formalizes the claim, tests it empirically, and derives its protocol and constitutional consequences.
That sentence appears in four registers across the corpus. In Volume I it appears as a historical observation: the merchant of Bruges, the notarial system, the bill of exchange — institutional machinery that has accumulated over thousands of years because someone has to pay the cost of making local truths compose into global ones. In The Proofs it appears as a cohomological obstruction, classified by a computable invariant. In Bulla it appears as a polynomial-time algorithm operating on tool schemas. In BABEL it appears as an empirical correlation between that algorithm's output and actual failure rates on 703 real compositions.
The page that follows does not argue the claim is true. That argument is made separately by the trilogy, the formal companion, the verification tool, and the benchmark. The page that follows argues instead that these are the same claim, not four claims that happen to share a phrase. Cross-register coherence is itself evidence: when one structure appears across four independent formalisms, each with different failure modes, the probability that the agreement is artifact drops sharply.
This is the across-register bridge. Its companion, Bridge Arguments, traces the across-volume bridge — how the trilogy's four equations compose into a single dependency chain from witnesses to mercy.
The four registers at a glance
| Register | What it says | Key artifact | Falsification |
|---|---|---|---|
| Philosophical | Coordination across contexts has an irreducible cost; institutions that charge a premium on top of that cost are extracting rent. | Vol I, especially the convergence thesis (Ch 1) and the sheaf-condition unification (bridge-arguments) | A durable coordination system that omits one of the five witness properties and survives at civilizational scale. |
| Formal | The obstruction to gluing local data into a global section is classified by a computable invariant — a cohomology class — under the sheaf condition. | The Proofs, especially A11 (sheaf condition), A17 (predicate invention), A21 (coherence cost model) | A bilateral-valid composition that fails globally with no identifiable cohomological obstruction. |
| Computational | The invariant reduces, under standard assumptions on tool schemas, to a rank difference: fee = rank(δ_full) − rank(δ_obs). Polynomial time from schemas alone. | Bulla, especially coboundary.py and witness_geometry.py; 14 Lean-verified theorems | An empirical mismatch rate uncorrelated with computed fee on a corpus of real compositions. |
| Empirical | The computed fee predicts real compositional failure, with near-zero false positives at fee = 0 and near-certain mismatch at fee ≥ 4. | BABEL, the 703-composition calibration (bulla/calibration), and the flagship benchmark family | Fee = 0 compositions that fail at non-trivial rates, or fee ≥ 4 compositions that pass at non-trivial rates, on adversarial test families. |
Four registers, one claim. The next section walks a single concept across all four, to show that the cross-register agreement is structural rather than nominal.
Walking one row: the coherence fee
The coherence fee is the clearest through-line: it is the only concept with a computationally identical definition in all four registers, carrying through without translation loss.
Philosophical register
In Volume I, the coherence fee appears as a historical observation. There is an irreducible cost to composing local truths into global ones: the merchant's word, adequate among his creditors in Bruges, does not compose across the bill of exchange's journey to Florence without a notary's seal, an endorsement chain, and a standing relationship between the two parties' banking agents. The seal, the chain, and the relationship are not free. Someone pays. The Codex introduction names this cost directly:
There is a real cost to composing local truth into global coherence — call it the coherence fee — and someone must pay it.
This is not yet a theorem. It is a claim about institutional history: that notarial infrastructure, guild chartering, double-entry bookkeeping, credit bureaus, and platform governance all exist, across cultures that did not exchange legal concepts, because coherence across contexts is costly, and societies have repeatedly rediscovered that someone has to pay. The five witness properties are the specification of what the coherence fee buys.
Formal register
The Proofs formalizes this observation. The sheaf condition (A11) states when local data glue into a global section. When the condition fails, the obstruction is a cohomology class: an element of H¹ whose non-vanishing classifies the failure. The coherence cost model (A21) proves that this obstruction increases monotonically under refinement of the cover. Finer composition is strictly more expensive, with the cost measured in cohomological terms.
This moves the claim from "someone must pay" to "the cost is computable, and grows in specific ways with specific inputs." The informal observation that bills of exchange required notarial seals becomes the formal claim that the obstruction to a global section is a rank-measurable invariant of the composition's topology. The philosophical register provides the historical weight. The formal register provides the mathematical structure that makes the historical pattern non-accidental.
Computational register
Bulla reduces the cohomology class, under standard assumptions on tool schemas, to a rank difference:
fee(G) = rank(δ_full) − rank(δ_obs)
Where δ_full is the coboundary operator over all fields including internal_state, and δ_obs is the coboundary restricted to observable fields. The difference counts independent semantic dimensions hidden from at least one tool but required by an edge. The computation is polynomial time in the composition graph, exact under rational arithmetic, and requires no execution. Schemas alone suffice.
The move from the formal register to the computational is not trivial. It required identifying the standard assumptions under which a cohomology class could be computed without constructing the full cohomology. The hierarchical fee theorem — fee(G) = Σ fee(Gᵢ) + boundary_fee(P) for any partition P — is the key result enabling polynomial-time decomposition. Fourteen Lean-verified theorems, zero sorry, establish the correctness of the reduction.
Empirical register
BABEL and the calibration corpus test whether the computed fee predicts actual failure. The registry calibration — 38 real MCP servers, 703 pairwise compositions — found Spearman ρ = 0.996 between boundary fee and actual mismatch rate, with zero false negatives among the 678 compositions where boundary fee = 0. A separate pilot on 10 servers in 45 pairwise compositions produced the zone table: fee = 0 safe, fee ≥ 4 unsafe at approximately 100% rate. On the synthetic scaling family (464 instances), the fee-based diagnostic achieves R² ≥ 0.86 against ground truth computed by symbolic execution.
The empirical register does not prove the formal or computational registers. It shows that the quantity they define is not an artifact. A reader who accepts the formal development but doubts its real-world applicability has 703 real MCP compositions to examine.
What the walk shows
Each register sharpens the prior without replacing it. The historical observation persists into the empirical row: the merchants of Bruges are not debunked by the Spearman correlation but exemplified by it. The four registers are not four arguments. They are four zooms on one structure that survives all four.
Gestured rows
Three other concepts traverse the four registers without requiring a full walk.
Witness structure
The five witness properties (binding, conditions, stakes, recourse, composition) recur across Vol I's convergent civilizations. The Proofs formalizes them as A1–A6: commitment, evidence, schema, epistemic status, coherence requirement, sense boundary. Bulla operationalizes them through content-addressed WitnessReceipts whose three hashes bind composition, measurement, and judgment separately. BABEL's bronze and silver real-MCP tracks test whether actual server ecosystems exhibit the same structure when measured on live endpoints. The register-crossings here are slightly looser than the coherence fee. Witness properties in Vol I are a specification. Witness receipts in Bulla are an implementation of that specification over a narrower domain — tool compositions rather than the general case of cross-institutional verification. Bulla's receipt format is an instance of the witness-property specification, not a universal one.
Boundary fee
Vol I identifies the intermediary's chokepoint as the site where the trust tax accrues: the notary, the bureau, the platform all sit at the boundary across which local truths must compose. The Proofs formalizes this through sheaf descent, where the global section's existence depends on agreement across the overlaps between local sites. Bulla's hierarchical fee theorem makes this computable: fee(G) = Σᵢ fee(Gᵢ) + boundary_fee(P) for any partition P, separating intra-component fee from boundary fee, which is the fee no individual component can see from inside itself. BABEL's cross-server tests (filesystem + GitHub, boundary fee = 1) demonstrate the theorem on live MCP servers: neither server can detect the path-convention mismatch from its own schema, but the composition's boundary fee identifies it exactly.
Predicate invention
Vol I's convergence thesis implies that new witness predicates — novel ways of naming what a witness attests — emerge under institutional need, not by fiat. The Proofs formalizes this as A17 (predicate invention) and A17b (conservative extension): a new predicate is admissible only if it is locally grounded, agrees on overlaps with existing predicates, and extends the existing theory conservatively. The Predicate Invention Under Sheaf Constraints (SCPI) paper extends this into a formal theory of descent-based predicate admissibility.
The empirical register is currently empty. No BABEL-equivalent benchmark measures predicate invention in the wild. This is the gap in the cross-register matrix, and it should be named rather than hedged: the philosophical, formal, and formal-adjacent registers converge on a claim that the empirical register has not yet tested. A future benchmark family — compositions in which a new predicate is introduced mid-pipeline and bilateral validity must be re-established — would close the row. Until then, predicate invention travels three registers, not four.
What cross-register coherence is evidence of
When one structure appears across four independent formalisms, each with different failure modes, the probability that the agreement is artifact drops sharply. Each register has its own failure mode — history could be wrong, formalism could be empty, algorithms could track nothing physical, correlations could be dataset artifacts — and the failure modes are independent. If the same structure survives all four, the conjunction is harder to explain as coincidence than as fact.
This is the methodological reason the ecosystem has four artifacts rather than one. Each register tests the others. Historical convergence, formal structure, algorithmic reduction, and empirical correlation do not merely translate one claim across media; they cross-validate a claim whose failure modes would otherwise be undetectable from inside any single register.
Vulnerability
The cross-register argument has specific failure conditions.
If the BABEL correlation degrades under adversarial or production-scale distributions, the empirical register's evidence for the computational reduction weakens. If the Proofs anchors do not fully resolve the Witness Claims in Vol I prose, the formal register's evidence for the historical observation weakens. If Bulla's standard assumptions on tool schemas fail on a nontrivial class of real compositions, the computational register detaches from the formal one. If the five witness properties fail to account for a durable coordination system that survives at civilizational scale without one of them, the philosophical register's foundation collapses.
Any one of these weakens the bridge without breaking the individual registers. All of them together would refute the claim that these are the same claim.