Most developers think about how to build software right. Security experts think about what can go wrong. Nir Valtman, co-founder and security veteran at Arnica, joins the show to talk about why those two mindsets are constantly in conflict, and what that means now that AI is writing your code. We get into prompt injection attacks, the real risks of agentic tools like OpenClaw, and a simple but powerful trick to make every AI coding session more secure by default. If you are building anything with AI tools right now, this one is worth your time.