#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI

The discussion centers on the current state and future trajectory of AI, particularly large language models. Key topics include the competition between US and Chinese AI companies, the debate between open-source and closed AI models, and the scaling laws governing AI development. The participants analyze the nuances of pre-training, mid-training, and post-training, as well as the role of RLHF and RLVR in enhancing model capabilities. The conversation also covers the potential for AI to transform various sectors, from software development to scientific research, while acknowledging the challenges of safety, bias, and economic disruption. The speakers also explore the potential for new architectures beyond transformers, such as text diffusion models, and the importance of human agency in guiding AI development.