Where's the voice cloning scam epidemic we were promised

The narrative two years ago was pretty clear: once voice synthesis got good enough, social engineering would explode. Executives would get spoofed into wire transfers. Kids would be catfished by synthetic voices of peers. The technology would democratize fraud in some unstoppable way. And yeah, the tech itself is genuinely good now. But the predicted wave hasn't materialized at anything like the scale people were warning about. There have been isolated incidents (the grandmother scam variant, a few CEO spoofing attempts that failed), but nothing approaching the "this is going to break trust in audio" scenario. I'm skeptical that the technical capability was ever the binding constraint here. The friction points are actually pretty boring: you still need someone's voice sample, you still need a pretext that lands, and you're still making a call to someone who can (and often will) verify through a different channel. The grandmother scam works because it's psychologically brutal, not because the voice quality is flawless. What interests me is whether we collectively overestimated how much pure technical capability drives social engineering, versus context and vulnerability. A convincing synthetic voice doesn't make someone's elderly parent more likely to believe a stranger—the social engineering still has to work. Maybe we were pattern-matching to "new technology = new attack surface" without checking whether the attacks actually scale.