MetaMask's phishing warning has been telling users not to proceed for years. People click through it anyway, often enough that the warning barely registers as a warning anymore, just a red screen between them and the thing they already decided to do.

That's the failure mode built into every warn-and-let-through system. A warning only works on someone who was already going to stop. Anyone who'd already decided just clicks past it, and after enough repetitions, the click becomes reflexive.

A policy that hard-blocks instead of warning removes that choice at the moment it matters, which sounds heavy-handed until you notice the warning was never really a choice for most people anyway, just friction they'd learned to ignore.

Newton's policy checks resolve to attestation or nothing, pass or the transaction doesn't proceed, not a red screen someone can click past. That's a narrower kind of safety than a system that tries to inform every possible user. It's also the kind that doesn't depend on someone actually reading the warning.

If a warning only stops the person who was already going to stop, was it ever protecting anyone else?

#newt $NEWT @NewtonProtocol