LWiAI Podcast #234 - Opus 4.6, GPT-5.3-Codex, Seedance 2.0, GLM-5
Summary
Hosts Andrei Kurenkov and Jeremy Harris discuss major AI model releases, including Anthropic’s Opus 4.6, OpenAI’s GPT-5.3 Codex, and Google’s Gemini 3 DeepThink. They analyze the shift toward parallel agent workflows, the hardware implications of the OpenAI-Cerebras partnership, and the blurring line between software and hardware. The episode also covers breakthroughs in video generation from ByteDance and significant funding rounds for companies like Eleven Labs.
Topics discussed
Intro, podcast updates, and disclaimer on politics
Anthropic Opus 4.6 and OpenAI Codex 5.3 releases
Recursive self-improvement and AI safety evals
Cerebras partnership and hardware efficiency
Gemini 3 DeepThink and ARC AGI benchmark results
ByteDance's Vio video model and capabilities
Jupoo AI GLM-5 and DeepSeq sparse attention
Cursor updates and XAI Grok Imagine API
Eleven Labs $11B valuation and Runway funding
Aptronic humanoid robotics and Mercedes pilots
Super Bowl AI ads and Anthropic leadership changes
SpaceX-XAI merger and Waymo production plans
Open source models and OpenClaw security risks
Tiny LoRA and Alpha World RL training paper
Vending bench benchmarks and task complexity trends
Anthropic research on misalignment and variance
Listen ad-free on Castria