Why an on-chain order book works here — when they usually don't The standard objection to on-chain CLOBs is that gas explodes twice: 1. Matching walks a sorted book — thousands of price levels, per-order state, O(n) loops over makers on every fill. 2. Makers requote constantly. On a CEX, the vast majority of messages are cancels. On-chain, every cancel is a transaction — and every stale quote is an MEV snack. Both problems come from one assumption: that makers must quote prices on a continuous grid. For options, you can factor the price: premium = BS(spot, strike, σ) Spot is observable — the Chainlink feed already knows it. Strikes are quantized — every contract lives on a fixed geometric grid (1.05ⁿ), one contract per rung, so liquidity can't fragment across arbitrary strikes. The only thing makers genuinely disagree on is σ — implied volatility. And a vol opinion doesn't need a continuous grid: 10 discrete IV levels cover it. So the "book" per contract collapses into 10 buckets and one bitmap word: ▸ Best quote = lowest set bit. No tree, no sorted list, no search. ▸ Within a bucket makers are fungible — one rebasing accumulator settles all of them. A fill is O(1) storage writes whether 1 maker or 1,000 are resting. ▸ Requoting doesn't exist. Premiums are computed from the live spot at fill time, so resting orders track the market automatically. The single biggest gas sink of traditional books — cancel/replace spam — is structurally gone. A taker sweep touches at most 10 buckets: ~300k gas, bounded, regardless of depth. The honest caveat: this isn't "CLOBs solved." It works because option premiums are model-priceable — you can push the continuous dimension (spot) into an oracle and quantize the opinion dimension (vol). Spot trading can't make that split. Derivatives can. http://roptions.exchange(Miki)