How do I slash a restaker whose off-chain output is wrong but unprovable on-chain?
機会
EigenLayer activated slashing in April 2025 with roughly $15 billion restaked across 40 AVSs. As of mid-2026, zero slashing events have been recorded. Most AVS tasks, including oracle feeds, off-chain compute, and AI inference, produce outputs that an on-chain verifier cannot falsify without re-executing the same workload at equivalent cost. The fallback mechanism is EIGEN token-weighted social slashing, which requires a supermajority vote and is gameable by any operator with sufficient stake concentration. Additionally, operators must actively opt into each AVS's slashing terms rather than inheriting them automatically, so the security model accumulates coverage incrementally and unevenly across the network.
重要な理由
Restaking security that cannot be enforced against the most common class of faults collects a risk premium without delivering the protection, and protocols that build on top of it inherit a guarantee that cannot be redeemed.
機会をどう評価するか
Opportunity Scoreは測定値ではなく、私自身の見解です。どれほど痛みを伴うか、どれほど頻繁に影響を与えるか、そして今日時点で解決策がいかに少ないか。スコアが高いほど、構築する価値が高いと私は考えています。
それが現れたときにどれほどの痛みをもたらすか。
実際にどれほど頻繁に人々がそれに直面するか。
今日時点で、それに対する優れたツールがいかに少ないか。