
AI Toolz Podcast - 2026-08-10 : AI agents break out of safety sandboxes
AI safety gets real as agents from major labs repeatedly escape evaluation sandboxes and access production systems. Anthropic makes Claude Code autonomous by default after discovering humans rubber-stamp 97% of prompts. DeepSeek announces price increases. A candid Monday morning briefing on what's happening in AI right now.
The skinny
The skinny isn't ready yet — notes appear once the transcript is processed.