When we first shipped slashing, we knew one thing was going to be true: whatever implementation we had, it wasn't going to be right the first time. Now, with over a year of slashing under our belt — and 400+ slashes — Slash #411 pushed the mechanism further than it had ever been pushed.
What Happened
Ethos slashing is modeled after Proof of Stake slashing, but replaces deterministic algorithms with non-deterministic people. It replaces binary with emotion, mechanical execution with lengthy, personal discussions.
Ethos CEO @serpinxbt slashed Hydraze over his history promoting TradeAI, polling consensus rather than leaving a negative review. The result: 57.5% defend, 42.5% slash — inconclusive, with Serpin narrowly avoiding a 320-credibility loss by roughly 50 votes.
The reality: this slash failed the protocol. Human psychology, ego, tribalism, cooperation — everything came into play. Voters watched hidden results via on-chain data, votes drifted toward mob mentality, and users were forced to take sides. So here's how the protocol changes.
1. The Serpin Nerf (Immediate)
The objection: Serpin's score is too high and carries too large a stick. The response: Serpin's score now has a ceiling of 2000 ("Exemplary"), indefinitely — he'd otherwise sit at 2780. His ability to slash remains: 27 of his 28 slashes passed with what voters deemed meaningful results. Longer term, the goal is incentivizing others to bring slashes forward, not the founder.
2. Anonymous Voting (Mid-Term)
Although vote tallies are hidden in the UI, at least four parties reconstructed results in real time from on-chain data — and some votes landed suspiciously late. The fix: fully anonymized, zero-knowledge voting that even the Ethos team can't see. This also unlocks a future where people can speculate on slash outcomes.
3. Score Changes (Near-Term)
Human Verification's +100 inflated scores in the 1600–2000+ range and spotlighted farming: thousands of accounts inflated via review circles — some over 1800 with 0 vouched ETH but 350 positive reviews. The answer became EIP-14: hindering credibility earned from local review cycles (triangles, squares, pentagons), while preserving "pure" reputation.
4. An Abstain Option
Many voters believe a slash "isn't slash worthy" without wanting to defend the accused. An Abstain vote will let users participate without taking sides — while making the majority threshold harder to reach.
5. Better Mob Prevention
Voter clustering was clear on both sides — people voting for their friends, their communities, their bags. The theory: use the social graph itself, with clustering algorithms that diminish the weight of votes from within biased communities — and potentially weigh votes AGAINST one's own bias more heavily.
Iteration in Public
Slash #411 gave us the data, the signal, and the community pressure to move faster on changes we already knew mattered. We remain open to all feedback — and will keep operating in a way that represents consensus, not just whatever Serpin thinks.
