In this TechTalk LIVE I go deep on why “agentic AI” is suddenly everywhere, and why it’s both exciting and kinda terrifying. I talk about the idea of having a little cluster of agents that behave like a mini dev team—planning, coding, iterating, and even developing their own slang and personalities. Then I show how tools like OpenCode can act like an AI shell: you can literally ask it to ping Google, traceroute Amazon, or reset your network card, and it’ll open PowerShell, run commands, read the output, and try again if it didn’t get the result you asked for. That’s awesome for productivity… and also a giant “do you really want this controlling your machine?” moment. Then I pivot into the wild west side of AI: image/video generation and how weak some guardrails are. I tested X’s Grok/Imagine hard, and I’m telling you right now—getting it to reject prompts is ridiculously difficult. It’ll edit real photos, it’ll make photorealistic deepfake-style content, and it’ll even generate convincing voice video from a text prompt. My big takeaway is simple: we’ve entered an era where basically anyone can be turned into content without consent, and that’s a massive societal and security problem—right alongside the layoffs and nonstop breaches we keep seeing week after week.