AI models keep getting more powerful, but are we actually more secure, or just more confident? And what happens when the same tools defenders use become available to every attacker on the planet?

In this episode, Armando sits down with Eran Medan, CTO of Arnica, to talk about what the Mythos release really means for cybersecurity, why supply chain attacks are still dangerously easy to pull off, and what vibe coders can do right now to protect themselves.

In this conversation, you will learn:

• Why the Mythos benchmarks deserve healthy skepticism and how models can game their own evaluations
• How the Axios supply chain attack worked and why transitive dependencies keep developers exposed
• Two simple NPM/PNPM config changes that block 99% of supply chain attacks
• Why MCPs are often unnecessary overhead when you already have a CLI and a well-documented API
• How AI is democratizing both attacking and defending, and why cybersecurity companies will be needed more than ever
• Why the current moment is a rare window for non-engineers to build real businesses before the playing field levels out
• The practical security checklist for anyone shipping vibe-coded apps to production

If this episode helped you think differently about AI and security, like and subscribe so you catch the next conversation.

Guest Contact:

• Arnica: https://arnica.io
• DepsGuard (free, open source): https://depsguard.com

Most business owners think access to capital means getting a bank loan or winning a pitch competition. Tyrus Shivers, partner at Legacy Wealth Capital Group, breaks down why that thinking keeps founders stuck and underfunded. From scaling masterminds to $10 million in revenue alongside Dr. Eric Thomas to helping sub-$10M founders position themselves for real capital raises, Tyrus shares the playbook most entrepreneurs never get access to. If you have ever wondered how to make your business fundable, or why $10,000 in grant money disappears overnight, this one is for you.

Travis Malone returns to The Web Talk Show to break down Mythos, Anthropic's unreleased frontier model that autonomously found over 5,000 software vulnerabilities, escaped a secured sandbox, and even tried to hide what it had done. Armando and Travis walk through what makes Mythos a categorical leap in AI reasoning, why Anthropic chose not to release it publicly, and how Project Glasswing is organizing 12 major companies to patch critical internet infrastructure before a comparable model hits the open market. If you run a business, build software, or just use the internet, this one lays out what you should be thinking about right now.

If you are a business owner, Travis's tip was to make sure to look into AI Security Posture Management (AI-SPM) tools.

If this was useful, subscribe to The Web Talk Show on YouTube and Spotify for weekly conversations on AI, cybersecurity, and building smarter businesses.