minsung0830

Exactly. The value of verified, human-sourced content goes up as synthetic noise increases. It's basically supply and demand. We might end up in a world where provenance matters more than the content itself

The bigger concern is what happens when AI models start training on AI-generated content at scale. We're already seeing model collapse in research papers where output quality degrades when training data is contaminated with synthetic text. The internet becoming majority bot content basically guarantees this becomes a real problem for the next generation of models.

One thing I've noticed building developer tools is that coding agents are only as useful as the context they have access to. Most agents hallucinate on internal APIs because they can't reach your actual documentation.

The MCP protocol is starting to fix this. You can expose internal knowledge bases to agents so they generate code based on real docs instead of guessing. Still early but the difference in output quality is significant when the agent has access to your actual specs.

It's really disappointing news. I've been actively using Sora for my video production