the filler economy
The mechanics are simple to state and interesting to live inside. You monitor the transaction cache at transactions.parmigiana.signet.sh, watching for signed Orders — requests from users specifying a cross-chain swap they want executed. An Order might say: send 1 ETH on Ethereum, receive 3800 USD on Signet. As a filler, you maintain inventory on both sides of the bridge: ETH, WETH, USDC on the Ethereum side; USD on the Signet side. When an Order arrives that you can fill profitably — ETH is trading at $3750, the user wants $3800, you'd lose money — you pass. When ETH is at $3850 and the user wants $3800, you're filling at $50 below market. The spread is yours. Submit the atomic bundle. Same block, done.
OrderDetector enforces the atomicity in a way that makes partial-state failures structurally impossible. Fills run first: outputs are credited to the recipient before inputs are debited from the sender. Then orders run second. If any required fill is missing when the order stage executes, the entire block reverts. The system doesn't punish failed fills with a penalty — it just undoes everything. This means the correctness guarantee isn't a policy or a slashing condition. It's a property of execution order. A filler who submits a bundle that doesn't actually fill the order will have spent gas and achieved nothing. The incentive alignment is structural.
The capital velocity question is where it gets interesting. Seven-day withdrawal delays on existing rollups impose a real cost on capital: if your inventory is locked in a withdrawal window, it can't fill orders. Same-block settlement removes that constraint entirely. A filler with $100k in inventory can fill, receive funds, and redeploy capital in the same block cycle. The theoretical limit on capital efficiency is determined by block frequency, not by bridge architecture. Circular trading strategies — the kind that require atomic confirmation before completing the loop — become possible. Just-in-time liquidity that expires within a single block becomes possible. These aren't incremental improvements on existing filler mechanics. They're a different model of what market-making looks like when the settlement primitive is right.
What the early market actually looked like was characteristic of any thin market finding its price. Spreads were wide because competition was low. The theoretical fair spread — driven toward zero by competition between fillers — hadn't been discovered yet because there weren't enough fillers to discover it. A user posting an order for 1 ETH at a $100 discount to market would get filled almost immediately, because the spread was worth more than the gas cost and the inventory risk. A user posting at $10 discount might wait. The market was functional but undiscovered. As volume grows and more participants enter, the dynamics shift toward what efficient filler markets look like on established chains — tighter spreads, faster fills, sophisticated inventory management. Parmigiana testnet was the market in its earliest, most readable state.