The cheapest route is not necessarily the correct route.
VŒID explores a decision layer for cross-rail money movement, where cost competes with timing, reliability, liquidity, operational posture, policy, and confidence.
OwnershipProduct model, deterministic Go prototype, builder and operator UX, workspace architecture, and system direction.
GODeterministic evaluatorTyped route request and response
RANK + FALLBACKDecision outputWinner plus ordered alternatives
LIVE / MOCKSource honestyRuntime references separated from fallback data
SHADOWExecution postureRecommendation without money movement
01
Context / stakes
Routing is a decision problem.
Cross-border and stablecoin workflows are fragmented across banks, mobile channels, treasury queues, OTC desks, and digital-asset rails. Each exposes a different fee, settlement time, failure surface, and operational dependency.
A cost-only router hides the actual decision. A slightly cheaper path can be wrong when liquidity is thin, a provider is degraded, settlement is uncertain, or policy requires a different counterparty posture.
02
System model
Normalize the route truth before ranking it.
Prototype decision flow
01RequestAmount, corridor, mode, markup
02CandidatesPrototype rail set
03NormalizeLatency, fee, and posture
04EvaluateDeterministic rank and fallbacks
05ExplainWinner, reason, and alternatives
06HandoffLicensed execution boundary
Implemented in the prototypePlanned boundary
SYS-02The request shape is part of the product.+
Amount and corridor are not enough. A useful decision contract also carries mode, constraints, markup, policy context, and the explanation the downstream operator will need.
03
Current system
What is implemented now.
Route worker
Deterministic ranking in Go.
The current worker accepts a route request, creates a prototype rail set, scores latency, fees, and rail posture, then returns a winner with ordered fallbacks.
Product shell
One workspace truth.
Next.js provides the public and product shell. Supabase supplies sessions, workspace setup, memberships, and row-level access policies.
Reference layer
Live data with visible fallback.
The dashboard can request FX and stablecoin references, but marks fallback values as mock when a provider is unavailable.
Operational rooms
Builder and operator views.
Separate rooms share the same route vocabulary: request packet, ranking result, fallback posture, and proof envelope.
Prototype interface / Builder roomA request becomes a stable, inspectable contract.
Implemented prototype
RequestCorridor · amount · mode
The evaluator receives explicit intent and constraints.
ResponseWinner · fallbacks
The result keeps alternatives alongside the first choice.
ContextDecision time · posture
Downstream operators receive explanation fields, not only an ID.
Decision provedThe Builder room proves that ranking is exposed through one explicit request and response vocabulary.
Prototype interface / Operator roomThe next valid route survives alongside the winner.
Shadow mode
RecommendedCandidate A
Highest-ranked path for the supplied constraints.
Fallback 01Candidate B
Ordered recovery path if the preferred rail degrades.
PostureRecommend only
The current system does not execute or custody funds.
Decision provedThe Operator room proves that recovery and human control are normal route state, not an afterthought. Candidate names are intentionally generic because production providers are not connected.
Prototype interface / Source stateReference provenance stays visible when live data falls back.
Implemented boundary
ReferenceLive
A runtime provider supplied the current value.
FallbackMock
A fixture is displayed only with an explicit source label.
Candidate factorsFee · timing · posture
Comparable factors feed the deterministic evaluator.
Decision provedThe source state proves that prototype fixtures are not presented as live market or traction evidence.
04
Planned / bounded
What remains direction, not traction.
Provider adaptersPlanned
Production rail availability, provider health, and normalized quote contracts are not yet connected.
Liquidity and reliabilityPlanned
The prototype does not yet ingest real corridor liquidity, settlement reliability, or counterparty telemetry.
Policy engineDirection
Policy-aware ranking is part of the thesis, but a production rule engine is not presented as complete.
ExecutionBoundary
Any eventual transfer belongs behind licensed partners, explicit controls, and an auditable handoff. VŒID is not presented as a custodian.
05
Decision doctrine
Determinism before irreversible action.
An AI layer can translate intent, investigate anomalies, and explain a recommendation. It should not become an unconstrained authority over a transfer.
Explicit inputs
Keep corridor, amount, mode, and constraints inspectable.
Stable rank
The same inputs and health snapshot should produce the same route order.
Fallbacks
A decision should preserve a practical second path, not only celebrate a winner.
Operator control
Overrides, holds, and execution handoff need durable policy and audit state.
OPS-01A fallback is part of the decision, not an error screen.+
Operational software should keep the next-best valid route visible before a provider fails. Recovery is designed into the normal state.
06
UX / design doctrine
Compact, smoked, and operational.
The current design direction uses compact chrome, smoked surfaces, restrained signal texture, and mono data treatment. Builder and operator rooms avoid oversized dashboard theatre because the route state is the product.
Status color is sparse. Dither and ASCII fields stay illustrative and bounded. Dense values use tabular numerics, while source pills make live and mock references visibly different.
07
Outcome
The honest outcome is a serious prototype.
A tested Go route worker with deterministic ranking and fallback output.
A Next.js product shell with Supabase-backed auth, workspace creation, and row policies.
Builder and operator surfaces built around a shared route vocabulary.
Live market references that degrade into clearly labeled mock state.
No invented customers, transaction volume, production providers, or custody claims.