The correspondent banking model has 6 toll booths between sender and receiver. How stablecoin rails change the architecture.

Cross-border settlement architecture hasn't changed much in decades. The standard model routes through sender bank, FX conversion, correspondent bank one, correspondent bank two, FX conversion again, receiving bank. Six independent parties, each extracting a fee.

The result: roughly 6.5% total cost on a standard transfer. $65 gone on $1,000 sent. Settlement takes 1-5 business days because each handoff adds queue time.

Stablecoin rails swap that six-party chain for a three-step model: on-ramp to stablecoin, network transfer, off-ramp to fiat. No correspondents, no clearing intermediaries, no redundant FX conversion. Cost drops to around 0.5%. Settlement time drops to seconds.

The architectural question this raises for anyone building payments infrastructure: where do the on/off-ramp layers live, and who holds the compliance surface for each? That's where a lot of the interesting infrastructure decisions are being made right now. Fintechs rebuilding their settlement layer aren't just picking a chain, they're also deciding how much of the KYC and licensing overhead they want to own versus delegate.

Disclosure requirements for local payment methods, cross-border licensing across different jurisdictions, and tiered KYC across the ramp layer are the parts of this that tend to get underestimated.

What part of this stack do people here think is still genuinely unsolved?