#253 - Opus 5, Gemini 3.6, Kimi K3, Hugging Face Hack
Summary
Hosts Andrey Kerenkov and Jeremy discuss recent AI developments, including Anthropic’s Opus 5, Google’s Gemini Flash models, and Black Forest Labs’ Flux 3. They analyze major funding rounds for Safe Superintelligence and Fireworks, alongside Anthropic’s compute partnerships with AMD and Meta. The episode also covers OpenAI’s ChatGPT Health rollout and policy debates regarding AI services provided to blacklisted Chinese entities via Singapore subsidiaries.
Topics discussed
Intro: Episode overview and host introduction
Sponsors: Factor Meals and Outshift by Cisco
New Models: Anthropic Opus 5, pricing, and Gemini Pro 3.1
Multimodal AI: Black Forest Labs Flux 3 and Meta's WhatsApp integration
Business & Compute: SSI/NVIDIA deal, AMD/Anthropic partnership, and Fireworks
Geopolitics: Export controls and Chinese tech subsidiaries
Open Source: Kimi K3, Thinking Machines, and benchmarking frameworks
Safety Incident: OpenAI rogue agent hack and the Killswitch Act
AI Risk: Pause AI movement and alignment challenges
Protests & Policy: AI pause march and OpenAI's safety principles
Regulation & Security: AI companion rules and quantum decryption risks
Recursive Self-Improvement: Weco.ai claims and industry skepticism
Listen ad-free on Castria