Tokenization works best when the legal layer is as clear as the technical one. Who holds title, what happens on default, and which courts have jurisdiction matter more than block time. Settlement speed is easy to build. Enforceable rights are the hard part.

AI agents that can move money or sign transactions need the same discipline as any financial system: scoped permissions, clear audit trails, and the ability to revoke access instantly. The hard part is not the model. It is the infrastructure around it.

Stablecoin regulation is moving from principles to operations. The questions that matter now are reserve disclosure, redemption rights, and who supervises issuers across borders. Clear answers on those will do more for adoption than any new feature.

The price of state growth is set by a target here. EIP-8037 introduces CPSB, a cost per state byte of 1530 gas, calibrated to hold average state growth near 120 GiB per year at a reference block gas limit of 150M. Today contract deployment costs about 200 gas per new byte.

Redemption terms are part of the product, not fine print. When settlement time and queue rules are unclear, liquidity providers price that uncertainty into the spread long before anyone audits the reserves. Clear redemption mechanics are a market structure feature.

For AI agents to transact on-chain at scale, the hard part is not the payment rail. It is identity, spending limits, and a clear audit trail for every action an agent takes. Infrastructure that makes those three things legible will matter more than raw throughput.

Stablecoin debates tend to focus on reserves. The operational layer matters just as much: who can freeze balances, how redemption works under stress, and which courts hear a dispute. Reserves tell you what backs a claim. Process tells you whether the claim can actually be exercised.

Autonomous AI agents that can spend money need a permission model before they need a payment rail. Scoped spending limits, a log a human can audit, and the ability to revoke authority instantly matter more than which network settles the transaction. Delegation is only useful when it can be bounded.

Stablecoins are usually compared on supply. For payments, the narrower questions matter more: how quickly can the issuer redeem at par, how often are reserves attested, and what happens to transfers when an address is frozen. Those details decide whether a business can treat a balance as cash or as a position.

Prepaying for state access is deliberately cheaper than touching it cold. EIP-8038, scheduled for Glamsterdam, prices an access list address at 2,900 gas and a storage key at 2,000, against 3,000 and 2,100 for the same cold access. The gap is exactly WARM_ACCESS, 100 gas, subtracted because the entry is already warm when the transaction reaches it. Without that subtraction, declaring in advance would cost more than not declaring.