Nine To Noon Tech: First Mythos - now the rest, RAMageddon and more
May 13, 2026
Mark Pesce, futurist and technology commentator who analyzes AI, cybersecurity and tech trends. He reacts to early Mythos claims and scepticism. He explains GPT-5.5 and rapid model escalation. He outlines rising AI-powered cyber threats and the Canvas ransomware impact. He describes RAM shortages from AI demand and odd emergent training behaviours like the “goblin” quirk.
AI Snips
Chapters
Transcript
Episode notes
AI Models Rapidly Escalate Cyber Capabilities
- Major new language models like Mythos and GPT‑5.5 rapidly match or exceed prior models in cybersecurity capabilities.
- OpenAI's GPT‑5.5 tested by the British AI Security Institute outperformed Mythos and shows this is a rising tide, not a one-off.
Cybersecurity Is Becoming AI Versus AI
- Cybersecurity is shifting to AI‑versus‑AI where attackers use AI to find exploits and defenders must use AI to patch them.
- Security teams are already discovering many more bugs in large codebases like Linux, Windows and macOS because of AI scanning.
Make Proactive Patching Nonnegotiable
- Treat proactive security as mandatory, not optional, because AI accelerates exploit discovery.
- The 'three‑month' urgency headline pushes organisations who were lax to prioritise patches and hardening immediately.