The Anti-SWIFT.
Hylaq Rail picks one rail per settlement. The Settlement Grid picks the cheapest path across rails, currencies, and geos — and commits the whole journey atomically. USD bank pay-in → EUR IBAN pay-out as a single quote. ~$5T/day moves across SWIFT alone, with 1–5 day settlement and 0.5–7% fees. The world pays a $300B/yr fragmentation tax. The Grid is how that ends.
The paradigm break
Money movement looks like a banking problem. It's actually a unit-conversion graph. Every payment is a path through nodes shaped like (currency, geo, custody) connected by edges that are either rails (FedNow, SEPA, SWIFT…) or FX swaps. The cheapest path is a Dijkstra query.
Until now nobody had identity (Hylaq), atomic settlement (Loadit), and an intelligent router (HQ) on the same plane. With those three, the graph becomes a one-line API: quote, execute, receipt. The Settlement Grid is that API.
What we replace
SWIFT correspondent banking
$24B/yr in fees on $5T/day
1970s-era message bus over 11,000 banks. 1–5 day settlement, 0.5–3% all-in. The Grid replaces the message + the routing + the FX in one atomic commit.
Cross-border FX spreads
$200B/yr extracted on retail + SMB
Banks quote 200–400 bps spread on retail FX. The Grid quotes one mid-market rate + 25 bps protocol spread. Hidden margin disappears.
Stablecoin bridges
$8B/yr extracted via slippage + bridge fees
Cross-chain bridges are slow, hackable, and routinely 0.3–1.5% all-in. The Grid treats CHAIN custody as just another node in the graph — same atomic guarantee, no bridge fees.
Multi-rail aggregators (Currencycloud, Wise B2B)
$6B/yr in API-fee margin
They sell access to the rails behind a black box. The Grid is open: every quote returns the full hop list with fees, latency, reliability — auditable.
Treasury rails-of-rails (BNY, Citi)
Tens of billions in custody + correspondent revenue
The Grid lets a corporate treasury query for the cheapest USD-EUR path live, without a banking relationship in either jurisdiction. Custody banks become commoditized.
Card interchange when cards aren't needed
$150B/yr globally
A retailer accepting cross-border online today eats 3–4% interchange. Recipient on Hylaq + Settlement Grid? FedNow → FX → SEPA at sub-cent total cost.
Five paths. One protocol.
Adaeze in Lagos, Berlin client
EUR SEPA pay-in → FX → USD HYLAQ → INTERNATIONAL_BANK to NGN. Today: 5 days + $35 + 4% FX. On the Grid: 12s, ~30 bps total. The Grid quote shows every hop.
Shopify merchant in Tokyo
Customer pays USD card-on-file. Grid routes CARD → USD HYLAQ → FX → JPY HYLAQ → INTERNATIONAL_BANK to merchant. 2.9% card rake compresses to ~50 bps protocol + processor.
200-person SaaS payroll
One USD treasury account → 47 employees in 12 countries. Grid quotes a fanout — atomic per-employee path, single audit-perfect signoff. Replaces the $200K/yr AP analyst.
Auditor on cross-border M&A
$2B acquisition with USD escrow → EUR closing. Grid receipt is one sha256 hash chaining every hop's per-rail proof. Auditor verifies offline. Replaces the 6-month confirmation dance.
DAO treasury rebalance
USDC chain → USD HYLAQ → FX → EUR HYLAQ → SEPA to the foundation's bank. Single quote, atomic execute, COMPENSATED rollback if any hop fails. No bridge dependency.
The Mesh paying its workers
HQ Compute Mesh is the Grid's first internal customer. Per-token client → worker payments settle through the Grid in HYLAQ-USD. Two trillion-dollar systems, one settlement layer.
How the protocol thinks
// One Grid settlement, four edges:
(USD, US, BANK) ──ACH──▶ (USD, US, HYLAQ)
(USD, US, HYLAQ) ──HYLAQ_DIRECT──▶ (USD, *, HYLAQ)
(USD, *, HYLAQ) ──FX (25 bps)──▶ (EUR, *, HYLAQ)
(EUR, *, HYLAQ) ──SEPA_INSTANT──▶ (EUR, DE, BANK)
// Total: ~$0.20 fixed + 25 bps spread + 5 bps protocol. ~12s.
The 40-year arc
- Years 1–3Diaspora corridorsOne corridor at a time (US ↔ MX, EU ↔ NG, UK ↔ IN). 30 bps total cost beats Wise. ~$5B/yr routed.
- Years 4–7SMB cross-borderStripe + Adyen become fallback rails. The Grid is the default destination for any merchant taking international payments. ~$300B/yr.
- Years 8–12Corporate treasuryCFOs query the Grid before they call BNY. Corporate cross-border treasury volume migrates. ~$10T/yr.
- Years 13–20Central banksBilateral CBDC settlement runs on the Grid. SWIFT replacement RFPs name it. ~$100T/yr.
- Years 20–40The plumbing of the worldThe Grid is invisible — it's how every cross-border payment moves. SWIFT is on life support. The $300B/yr fragmentation tax is gone.
Shipped today
- src/lib/settlement-grid/ — types, fx-rates (18 currencies), corridors graph, Dijkstra pathfinder, quote builder, in-memory liquidity meter, atomic execute with commit/rollback, multi-hop signed receipts, quote cache, BigInt-safe wire serialize
- 3 API routes — POST /api/settlement/quote, POST /api/settlement/execute, GET /api/settlement/health
- Multi-hop pathfinder — Dijkstra over (currency × geo × custody) graph, ranked by USD-cents cost + reliability + latency
- Atomic execution — commit per hop with stub adapters; on failure, compensates every prior hop in reverse and unwinds liquidity reservations
- Hash-chained multi-hop receipt — sha256(canonicalize(header) || hashChain(outcomes)); verifyReceipt() works offline
- Jest unit tests — same-currency same-geo, same-currency cross-geo, cross-currency with FX, end-to-end settle + verify, COMPENSATED on hop failure
- Phase-2 swap points called out — replace USD_RATES with live provider, swap stub adapters for real per-rail providers (Increase, Stripe, Wise), persist to Postgres, replace meter with Redis
One quote. One execute. One receipt. Anywhere on Earth.
The world burns $300B/yr moving its own money around. The Settlement Grid replaces that with a single API call — and the HQ Compute Mesh is already its first internal customer.