Airgap closure, Part 4. Follows “Extraction is conserved: from MEV to GEV.”
In Part 3 the framing got corrected in the replies, and the correction matters enough to lead with it. The conserved quantity is not extraction across channels. It is extraction across preconditions. A channel is a surface. A precondition is the structural fact that makes the surface extractable in the first place. Sequencing privilege is one precondition, and it surfaces as both ordering extraction (MEV proper) and liquidation-ordering extraction. Close the surface while the precondition stays open and the value reappears somewhere else with access to the same privilege. Close the precondition and every surface that depended on it closes at once.
Seven channels, four preconditions
This reframing shrinks the target. There are seven channels in the original list, but they do not rest on seven independent foundations. Mapped down to preconditions, the set is smaller:
- Ordering discretion. Some party decides the sequence in which pending operations apply. Surfaces: transaction-ordering MEV, liquidation ordering.
- Parameter discretion. Some party can move protocol parameters by privileged vote or admin key. Surface: governance extraction.
- Mandatory intermediation. Use of the protocol forces a holding or a fee through an instrument that captures rent. Surface: token rent-seeking.
- Allocation asymmetry. Some party prices or allocates with information or access the counterparty does not have. Surfaces: insider capital-formation discounts, oracle rent, platform data monetization.
Four preconditions, not seven channels. The closure program is therefore: remove each precondition with a mechanism that leaves no privileged residue, and verify that nothing relocated.
The honest test for a closure is falsifiable in the direction that hurts:
Show a mechanism that removes a precondition and yet GEV is unchanged because the extracted value moved to another surface of the same precondition, and the closure was cosmetic. Show a surface closure that leaves its precondition fully open yet reduces total GEV with zero relocation, and conservation itself is wrong.
I want both tests on the table, because a conservation claim that cannot be broken is not doing work.
Precondition 1: ordering discretion
This post closes precondition 1: ordering discretion.
The standard MEV mitigations attack the symptom. Encrypted mempools hide order content until inclusion. Solver and builder auctions sell the ordering right to the highest bidder and rebate part of the proceeds. Private relays route around the public surface. Each of these reduces measured MEV. None of them removes ordering discretion. An encrypted mempool still has a sequencer who decides order once decryption happens. A builder auction does not abolish the ordering right, it formalizes its sale. The precondition is intact, so the extraction is conserved: it shows up as builder profit, as relay exclusivity, as the liquidation surface that the same privilege still feeds.
Removing the precondition: commit-reveal batch auction
Removing the precondition means there is no ordering right to hold, sell, or rebate. A commit-reveal batch auction with a uniform clearing price does this directly, and it is the mechanism I have implemented in VibeSwap, so I can describe it concretely rather than abstractly.
Operations in a fixed window are collected, not sequenced. In the commit phase each participant submits a hash of their order plus a secret, with a deposit. No content is visible, so there is nothing to order against. In the reveal phase participants open their commitments. Failure to reveal a committed order is slashable, which removes the option of committing widely and revealing selectively. Settlement then does two things that, together, delete the precondition rather than reassign it.
First, intra-batch order is not chosen by any party. The settlement order is a deterministic shuffle (Fisher-Yates) seeded by the XOR of the revealed secrets. No single participant, and no sequencer, controls the seed, because the seed is a function of everyone’s secret at once. There is no position in the sequence that anyone can buy, because the sequence is not for sale to begin with. It is not produced by discretion.
Second, every order in the batch that trades does so at a single uniform clearing price. Within a batch there is no better and worse position in line, because price does not depend on position in line. The thing MEV extracts, the spread between where you would have cleared and where a privileged reorder makes you clear, has no place to live. There is no privileged reorder, and there is no per-order price to improve by moving you.
Why this respects conservation
The reason this respects conservation is the part worth stating carefully. Batch auctions do not push ordering extraction into a builder market or a relay. They remove the privilege the builder market was pricing. Because the surface and the precondition coincide here, closing the precondition closes both surfaces that rode it. The liquidation-ordering surface dies for the same reason the transaction-ordering surface dies: there is no ordering discretion left to grant a liquidator priority.
If anyone can exhibit value extracted from intra-batch ordering after this, with the seed actually a function of all secrets and the clearing price actually uniform, the claim is false and I want to see it.
What this does not do
What this does not do is also the point. Closing precondition 1 leaves preconditions 2, 3, and 4 fully open. A protocol with perfect batch settlement can still extract through parameter discretion, through mandatory token intermediation, and through allocation asymmetry. This is exactly the conservation thesis applied to my own solution: a partial closure that stopped here would let extraction relocate to the open preconditions, and the GEV total would barely move. That is why the airgap does not close until the preconditions close together. Uniform clearing is necessary and visibly insufficient.
Next: precondition 2, parameter discretion
Part 5 takes the next precondition: parameter discretion. The closure there is not a different auction, it is a different category of fix. You cannot batch your way out of governance capture. The tool is a math-enforced invariant that bounds what any vote is allowed to do, so that the parameter is constrained by structure rather than protected by the good behavior of a majority. Augmented governance, stated as a precondition closure, is the subject of the next post.
A note on what I am claiming
The batch-auction mechanism described here is implemented, not yet proven in adversarial production at scale, and I will keep the demonstrated and the designed separate as the series continues. The conservation framework is a lens, and a lens earns its place by making falsifiable predictions about where value goes when you push on it. So far, every partial MEV fix I can find relocates rather than removes, which is what the lens predicts. The constructive claim is that removing the precondition, not policing the surface, is the move that does not relocat