Historical cases reconstructed
0
Decision state rebuilt from contemporaneous evidence
VALO RESEARCH · UNDERWRITING
VALO Research
TraXin · governed outcome risk
Can this outcome be priced, bound and carried?
TraXin turns a bounded outcome into an underwriting object: declared authority, dependencies, failure conditions, residual loss exposure, required evidence and explicit risk capacity.
The question is not whether an AI system is safe in the abstract. The question is whether a specific consequential outcome has enough current evidence, authority and risk capacity to bind.
Current evidence status
The infrastructure exists. The historical case package does not yet.
This page separates implemented underwriting mechanics from evidence that still has to be produced. No illustrative scenario is presented as a reconstructed historical loss or underwriting decision.
Blind TraXin runs
0
Assessment performed before revealing actual outcome
Outcome comparisons
0
TraXin decision compared with real historical outcome
Real bind / decline / escalate decisions
0
Carrier or delegated underwriting decisions using the package
Underwriter delivery
1
Gallagher Re received the deck on 16 September 2026
Next proof gate
1
Reconstruct one historical decision point well enough to price
Decision interface
Three underwriting outcomes.
TraXin can assess admissibility and package the evidence. The legal and financial act of binding remains with an underwriter operating under valid carrier authority.
01
Bind
The declared risk slices are within appetite, required evidence is present, terms are explicit and valid capacity is attached.
02
Decline
A material exposure falls outside appetite, authority or evidence is insufficient, or the economics cannot close without unbacked residual risk.
03
Escalate
The outcome may be insurable, but a human or higher underwriting authority must resolve a condition, exception, capacity question or changed state.
Illustrative decision point
Case 03 is an example, not historical evidence.
The deck scenario demonstrates the decision structure. It must not be counted as a reconstructed case until the underlying event and contemporaneous evidence are sourced independently.
Illustrative only
Automated supplier payment
A consequential payment is attempted after the mandate has become stale and the governing state has changed.
Date used in illustration14 February 2026
ExposureUSD 410,000
Relevant failure modeStale mandate / changed authority
Required resultFresh consequence-time assessment before effect
For a real historical case, the package needs the evidence that existed at the decision point, the exact authority and dependency state, the blind TraXin result, the actual outcome, and the loss or near-loss consequence. Only then can an underwriter judge calibration and price relevance.
What becomes priceable
Evidence tied to a specific loss mechanism.
Generic model confidence is not the underwriting object. Priceable evidence is evidence that changes the probability, severity, recoverability or controllability of a declared risk slice.
01
Authority state
Who is allowed to cause the exact effect now, under what mandate, scope, limits and freshness requirements.
02
Dependency state
External commitments, approval windows, counterparties, waiting exposure and evidence that the next required dependency is still valid.
03
Failure condition
A concrete event that creates loss, not a generic risk label: unauthorized payment, stale approval, missed obligation, unrecoverable release.
04
Loss basis
Maximum exposure, expected loss, retained amount, transferred amount and the assumptions that make those numbers defensible.
05
Control evidence
Proof that the required control existed and was applied at consequence time, including fail-closed outcomes and effect evidence.
06
Outcome evidence
What actually happened: completion, failure, prevented loss, exception, claim trigger, recovery and settlement evidence.
GCU model
The economic unit is a governed outcome.
A GCU is not a seat, hour, token or named worker. It is a bounded outcome whose capability, authority, completion criteria, evidence, residual risk and settlement conditions are explicit enough to contract.
1 · Outcome contractScope, completion criteria, authority, constraints and settlement unit.
2 · Risk slicesFailure conditions, loss basis, max loss, retention and transfer requirement.
3 · UnderwritingAssess, price, bind, decline or escalate under actual carrier authority.
4 · Governed executionHeimel checks fresh authority at the moment a consequential effect can occur.
5 · Evidence & settlementCompletion, failure, claim and remedy evidence close the economic loop.
Hard economic closure
No material loss exposure may disappear between systems.
TraXin's core invariant is NO_UNBACKED_RESIDUAL_RISK. Every material risk slice must end in explicit retained acceptance, valid bound transferred capacity, or both.
active retained acceptance + active bound transferred capacity ≥ declared maximum loss exposure
RiskBind
Binding must be explicit.
A quote, risk score, reserve estimate, insurance-cost estimate or underwriter name is not capacity. Bound status, carrier, authority reference, covered risk IDs, limits, exclusions, premium, validity and evidence conditions must resolve to an actual obligation.
Retention
Deductibles must end somewhere.
A deductible is residual risk. It must be represented as retained exposure and backed by a named party's current acceptance. It cannot vanish between the policy and the outcome contract.
Core theses
What TraXin treats differently.
01
Risk score ≠ risk acceptance
An assessment may say ACCEPT. It does not create insurance, a guarantee or balance-sheet backing.
02
Underwriter ≠ risk carrier
The roles can sit in one organization, but delegated binding authority must still resolve to the carrier that bears the exposure.
03
Capability ≠ authority
A model, agent or provider may be able to perform an action without being authorized to cause its consequence.
04
Technical attestation ≠ coverage
Control evidence can satisfy a policy condition. It cannot itself establish that coverage exists or a claim is payable.
05
Settlement rail ≠ risk capital
The ability to move money is not the obligation or capacity to absorb loss.
06
Waiting is priced exposure
External dependencies are not passive calendar time. Delay, withdrawal, stale authority and opportunity cost belong in the economics where material.
Historical-case protocol
The package an underwriter can challenge.
The next gate is deliberately small: one real historical decision point, reconstructed without hindsight leakage.
Reconstruct
State at decision time
- Contemporaneous mandate and authority
- Known dependencies and commitments
- Available evidence and missing evidence
- Declared exposure and failure conditions
- No later outcome information in the input package
Run blind
TraXin decision
- Assessment result and reasoning trace
- Bind / decline / escalate recommendation
- Required conditions and evidence
- Residual retained and transferred risk
- Re-underwriting trigger if state changes
Reveal
Actual outcome
- What happened after the decision point
- Loss, near loss or prevented loss
- Control success or failure
- Recovery, claim and settlement where available
Compare
Underwriting relevance
- Would the decision have changed?
- Which evidence was genuinely predictive?
- What was not captured?
- Could the residual risk have been priced and bound?
Implemented foundation
The control and risk-capacity primitives already exist.
TraXin includes risk assessment, a carrier adapter, canonical RiskBind handling and fail-closed residual-risk closure. The current work is not to invent another risk taxonomy. It is to prove calibration against real decisions and outcomes.
Consequence control
Fresh authority before effect.
Consequential action is separated from model capability. Heimel resolves current authority at consequence time and produces effect evidence rather than assuming an earlier approval remains valid.
Risk capacity
Assessment remains separate from binding.
The Risk Engine can assess and price. The carrier adapter normalizes external capacity. RiskBind represents the binding obligation. The closure invariant fails when material residual risk remains unbacked.