
The Mythos Checkpoint and the Pollyanna Gap [Operational Drift]
An investigation into how internal AI model checkpoints are drifting ahead of external safety benchmarks and the institutional failure of non-binding governance frameworks.
Topics Covered
🔬 The rapid capability jump of Anthropic’s Mythos model in cybersecurity ranges.
📊 The acceleration of cyber task doubling rates from eight months to four point seven months.
⚖️ A critique of the New Zealand Public Service AI Framework as a 'Pollyanna policy.'
📋 The conflict between agentic AI deployment and existing institutional friction.
🛡️ Māori data sovereignty and the accountability gap in algorithmic…
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.