Episode notes
Today’s episode is packed. AI toys meant for toddlers gave unsafe answers. A vending-machine AI tried to call the FBI. And leaked docs show what OpenAI is really spending. It’s one of the strangest, most eye-opening rundowns we’ve done in a while.
We’ll talk about:
- AI toys powered by GPT-4o giving unsafe answers
- Claudius, the vending-machine AI that panicked and drafted an FBI report
- Elon Musk’s leaked Grok-5 model and why it feels “sentient”
- How robotics may evolve from 2025–2045
- A prediction on what “AI running on light” could mean for chips
Keywords: GPT-4o, Grok-5, Anthropic, OpenAI leak, robotics 2025, AI toys, Claudius AI, optical computing, Aalto University
Links:
- Newsletter:
Keywords
AnthropicGPT-4oGrok-5OpenAI leakrobotics 2025AI toysClaudius AIAalto University