Kimi K3 Is Here. China Just Hit the AI Frontier.
Summary
This episode covers China’s Kimi K3 model nearing US frontier capabilities, leaked details on Anthropic’s upcoming Opus 5, and Demis Hassabis’s call for AI self-regulation. The hosts also discuss OpenAI’s rumored smart speaker and coding keyboard, the Suno AI hack revealing training data sources, and viral robot fighting footage. Finally, they showcase personal projects, including a browser-based camouflage game and an autonomous AI-generated TV show featuring a fig and a tardigrade.
Topics discussed
Intro: Kimi K3, AI games, and show welcome
China's Kimi K3 model and Anthropic usage limits
Kimi K3 benchmarks and progress vs US models
Alpha leak: Opus 5 testing and efficiency
AI regulation framework and GPT-5.6 Sol Ultra
OpenAI's portable smart speaker and privacy
Noise-canceling headphones and show feedback
Suno AI training data and copyright issues
Robot fighting league and floating bots
AI-generated 3D scenes in Blender
Camouflage cursor game built with AI
Fig and Moss: AI agents creating their own world
Listen ad-free on Castria