What you'll learn

• 1 (01:33) **Setup: The Late-Night Panic Over a Leaked AI** - YouTuber Matthew Berman records an urgent video from a disconnected vacation, driven by unease over a leaked AI model called Mythos.
• 2 (03:24) **The Scale: 10 Trillion Parameters vs. 1–2 Trillion Norm** - Mythos operates at a scale that doesn’t just improve chatbots but crosses into anomalous emergent behaviors.
• 3 (04:03) **The Bizarre Response: Bottling It Up Instead of Releasing It** - Anthropic does the opposite of standard Silicon Valley hype: they actively suppress Mythos and form a secret defensive coalition.
• 4 (05:55) **Mutually Assured Destruction: Why Rivals Collaborate** - The coalition isn’t altruism; it’s a survival pact against an existential digital threat.
• 5 (08:08) **AI Just Ate Software: The Zero-Day Discovery** - Mythos systematically dismantles decades of human engineering by autonomously finding thousands of zero-day vulnerabilities.
• 6 (10:15) **Vulnerability Chaining: The Real Threat Vector** - The most alarming capability isn’t finding single bugs but autonomously chaining them into catastrophic breaches.
• 7 (12:02) **The Patches Are Flawless: A Fundamental Shift in Logic** - The human maintainers on the receiving end noted that Mythos’s fixes were indistinguishable from code by highly skilled senior engineers.

+ Full timestamped outline available in the app