Pipe
A stablecoin payments product for Indonesia: pay any QRIS merchant through a licensed-bank rail. Built solo, in Antler's residency.
The problem
The wallet is the easy 10%. The hard, defensible 90% is the rail that makes paying a real merchant in rupiah with offshore stablecoins both legal and instant.
Indonesians who hold stablecoins can't spend them where they actually shop. QRIS, the national QR standard, is on every warung counter and every merchant in the country. Stablecoins are nowhere near it. The gap between offshore crypto and onshore rupiah is technical, regulatory, and trust, all at once.
Pipe closes it: hold USDC or USDT, scan any QRIS code, and the merchant is paid in rupiah, end to end in under a minute. The user never opens a bank account and never sees the machinery.
I designed and built Pipe largely on my own, in Antler's residency, working directly with the bank, the custody provider, and the regulators who make a rail like this legal.
Decisions
Make the licensed-bank rail the moat, and build it first
Anyone can ship a wallet screen. The defensible part is the rail underneath: a distributed saga that locks a quote, settles the stablecoin on-chain, and pays the merchant in rupiah through a Bank Indonesia-licensed bank, reconciled if any step fails. I built that first, with the bank's QRIS rail on one side and MPC custody plus on-chain settlement on the other. The hardest, most regulated part is exactly the part worth doing: the B2B2C merchant arrangement and the licensed rail are roughly a year to replicate, and that year is the moat.
Settle, never speculate
Pipe is a settlement business, not a trading book. It holds the aggregate rupiah liability and takes a clean spread on the conversion; it never bets on the price of the asset it moves. That is what makes it bank-grade rather than another crypto app, and it is the difference between a durable rail and a fragile one.
Hide every gram of crypto from the merchant, and most of it from the user
The merchant should never know crypto was involved. They get a normal QRIS payment in rupiah. The user shouldn't fight bank KYC, OTPs, or PINs at the counter either. Pipe absorbs the custody, settlement, and FX complexity, so the visible product is just: scan, confirm, done. Adoption lives or dies on the boring side being invisible.
Treat sub-60-second settlement as the real product constraint
Paying at a counter is a moment with no patience in it. The hard engineering goal was an end-to-end target under 60 seconds, across an offshore chain and an onshore bank, with clear states when a leg is slow. I designed the orchestration and the failure handling around that single number, because everything else is irrelevant if the person at the till is left waiting.
Build it hands-on to find the constraints specs hide
I built the mobile app and the settlement backend myself. Building it, rather than writing a spec, is how the real constraints surfaced: settlement timing, the exact failure states, and the one line of copy a first-time user needs at the moment they part with money.
Where it stands
Pipe is early, and that is the point. The interesting thing is not volume yet, it is that the hard regulated rail (a licensed bank, a real corporate rupiah account, on-chain settlement) exists and works end to end. The defensible work is done.
Mobile and web apps are built and in active development, with app-store submission underway. The product is Antler-associated, and the regulated rail runs on a Bank Indonesia-licensed partner, not an API bought off the shelf.