Anthropic just built an AI so “dangerously good” at hacking that they’ve locked it in a digital vault—and we’re cracking it open to uncover the Claude Mythos behaviors that terrified its own creators.
LINKS MENTIONED IN THIS EPISODE
AI Models & Technical Reports
Anthropic – Claude Mythos Preview System Card
If Anyone Builds It, Everyone Dies by Eliezer Yudkowsky
The Revenge of Analog: Real Things and Why They Matter by David Sax
Nicholas Carlini’s Analysis of Claude’s Hacking Capabilities
Angine de Poitrine (Microtonal Math Rock) | Full Performance
PODCAST INFO:
Podcast Website:
UpAgainstReality.com
Episode link:
https://upagainstreality.com/2026/04/14/claude-mythos/
Apple Podcasts:
podcasts.apple.com/us/podcast/up-against-reality/id1689699654
Spotify:
open.spotify.com/show/29wpR2Uf0d19FjTH2QubcD?si=a6359f9aa1964170
RSS:
upagainstreality.com/feed/podcast
Instagram:
https://www.instagram.com/upagainstreality
