Now with More Post Mythos Apocalypse!

In this episode of Marketing Over Coffee, Christopher Penn and John Wall dive into the buzz around Anthropic's new 'Mythos' AI model, which claims to autonomously hack systems by looping through tasks until success criteria are met. While acknowledging the hype, they emphasize that looping LLMs—like GLM 5.1 and Opus 4.6—are already capable of long-horizon attacks, raising serious cybersecurity concerns, especially for outdated embedded systems in critical infrastructure like power grids and water valves. The hosts stress that while AI-powered hacking is a real threat, the biggest vulnerabilities remain low-hanging fruit: weak passwords and unpatched legacy hardware. They also discuss Google’s return of Data Studio, the launch of Gemma 4—an open-weight, agentic AI model capable of running locally on devices—and the growing importance of local AI for privacy. The episode highlights practical tools like Framer for rapid website building and NerdWallet/Fundera for small business financing. A lighthearted segment critiques misleading marketing surveys and brand text fatigue, with John sharing his zero-tolerance policy for brand SMS. The hosts wrap up with personal updates on media consumption, including a glowing review of *Mad Max: Fury Road* and a DIY YouTube download system using Cloud Code and Jellyfin. Key takeaways include: 1) AI-powered hacking is real and evolving rapidly, especially through looping LLMs; 2) Patching basic vulnerabilities like weak passwords and outdated systems is more critical than ever; 3) Local AI models like Gemma 4 offer privacy and performance advantages; 4) Tools like Framer can drastically reduce website development friction; 5) Avoid overpaying for streaming services—use pulse subscriptions instead; 6) Be skeptical of marketing studies lacking transparency in methodology or sample size; 7) Use automation tools to manage media downloads without paying for premium subscriptions; 8) Prioritize privacy by running AI locally when possible.