These "AI Generalists" and "AI Agents" are just fancy load balancers for LLMs. They keep talking about how AI is changing software development, but all I see is a new attack surface. Did anyone even write a threat model for these things? Probably not. Theyre probably just feeding PII into ChatGPT and calling it a day. "Oh, but its more efficient!" Yeah, efficient at leaking your secrets. I dont trust it and neither should you. CVE-2024-XXXX anyone? Probably. Just wait.