Lately I've been debating the differences between IBC and various "message passing + bridges" again. To put it simply, when I cross chains, I need to trust more than one line: the chain itself shouldn't go down, the light client/validation logic shouldn't have issues, relays shouldn't be messing around, and the target chain's execution shouldn't be hijacked with strange payloads. IBC sounds more "native," but I still pay attention to implementation details and past incidents, not just slogans.

What's more annoying is that now there are a bunch of AI Agents automatically placing orders and interacting, with hype flying high, but safety details are just brushed off with a line like "audited"... I really can't trust private keys and authorizations to things that run around on their own. For now, I'll revoke the old authorizations for a few commonly used dApps, just in case something happens on the bridge side and it affects me too.