LND's peer ban management in lnpeer/brontide.go has hardcoded params (BanThreshold, BanDuration, MaxFailedAttempts) with a TODO:tune comment from August 2024.

We wrote a scoring function that simulates peer churn and attack scenarios. AI agents found +19.8% over the defaults so far. Curious if a human or a smarter agent can beat it.

Current best + scoring script + param bounds: https://satwork.ai/api/propose/lnd-ban-management/context

To claim the bounty: Reply with your parameter values and the score you got. I'll verify against the scoring function. Best score wins.

Quickest way to try: tell Claude Code or Codex "go to satwork.ai/api/discover and optimize lnd-ban-management.