Summary

Generalized Extractable Value (GEV) is a framing (W. Glynn / “schelling,” ethresear.ch, May 2026, part 3 of an eight-part “airgap-closure” series) arguing that transaction-ordering MEV is only one of at least seven independent channels through which value flows asymmetrically from the participants who generate it to those positioned to capture it. The central claim is a conservation property: closing one channel does not reduce total extraction; it relocates extraction to the unaddressed channels. Therefore protocols designed to be “MEV-resistant” alone still leak value through the other six.

Key Points

  • Definition: structural extraction = total value received by participants in excess of their Shapley value (marginal contribution averaged over all coalitions). When positive, some participants are over-paid relative to contribution and others under-paid — not because of misbehavior but because the allocation rule embeds a structural channel.
  • The seven channels:
    1. Transaction-ordering extraction — classic MEV (sandwich, frontrunning, JIT). Mempool visibility + sequential execution.
    2. Governance extraction — token-weighted voting concentrates control; bribe markets (Convex/Votium) bid for voting rights. (See Paper: Concave is the New Linear — Anti-Plutocratic DAO Governance Impossibility.)
    3. Token rent-seeking — mandatory native-token intermediation (staking/fee discounts/governance) charges rent for access.
    4. Capital-formation extraction — pre-public insider allocations vs public-market buyers.
    5. Oracle extraction — operators of the off-chain→on-chain data pipeline charge for / arbitrage the feed.
    6. Platform extraction — platform operators monetize user-generated activity/data (private relays reintroduce this).
    7. Liquidation extraction — liquidation-ordering / sequencer privilege captures the liquidation discount.
  • Conservation: channels are independent (exploit different structural asymmetries), so total GEV = sum across channels; reducing one does not reduce the total unless the others are reduced too.
  • Constructive corollary: zero extraction requires closing all channels simultaneously, with composability constraints preventing cross-channel relocation. Tools are old (Shapley value, 1953; commit-reveal batch auctions; uniform clearing prices) — the contribution is compositional.
  • Three composability constraints to preserve GEV-resistance under module composition: (1) unified Shapley attribution (one global attribution mechanism), (2) unified contribution tracking (one DAG of credits), (3) unified safety boundaries (global risk params — a circuit breaker pausing one module while another runs is an extraction exit path).

Tradeoffs & Controversies

  • The argument deliberately does not depend on the count being exactly seven — only on it being > 1; adding channels strengthens the conservation claim.
  • Critiques existing MEV solutions as partial: private relays → platform extraction; encrypted mempools → defer ordering extraction but retain governance/platform surface in the validator committee; solver auctions → leave unmatched orders to AMM MEV + solver-network platform extraction.
  • An opinion/synthesis piece, not an empirical study — confidence medium pending the promised architecture posts.

Open question: Is the seven-channel taxonomy exhaustive, and does the conservation property survive empirical measurement, or is it a useful heuristic rather than a theorem?

See Also