I once thought anti-Sybil was a problem of recognition. Who is real, who is a bot, but the longer I looked, the more I felt something was off. Systems keep trying to "snapshot" identity at a moment in time while behavior always flows.
The issue is not a lack of data but how we frame it. KYC, social graph, wallet history… all of these are snapshots, and snapshots are easily misinterpreted. Real people can look like bots while bots increasingly resemble humans.
SIGN, at least from the way I observe, seems to be trying a different approach. Not attempting to answer "who are you" but rather "what have you done and who verifies that." A layer of attestation where behavior is recorded as pieces of evidence with context, signed by individuals. It sounds simple, but it changes how systems think: from identity to credibility.
This does not eliminate Sybil; it just makes simulation more costly and less natural over time, and that seems to be the crux.
I remain skeptical because everything will become clear when there are real incentives, and someone intentionally disrupts, but if I have to choose a direction to follow, this is the direction I’m looking at; I will continue to observe further...