Insights from the Lex Fridman Podcast episode “#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI”, published February 1, 2026.
In "#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI", Refers to the January 2025 release of DeepSeek R1, an open-weight Chinese model that rivaled top US proprietary models at a fraction of the training cost. It signifies a geopolitical shift where the "moat" of massive compute…
In "#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI", A post-training technique where models are trained on tasks with objective ground truths (math, code) rather than subjective human preference. By grading the model purely on accuracy, it allows for massive scaling of reinforcement…
In "#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI", The process of allowing a model to generate more tokens (hidden "thoughts") to reason through a problem before outputting the final answer. Unlike pre-training scaling which is a fixed cost, this shifts the compute load to the…
In "#490 – State of AI in 2026: LLMs, Coding, Scaling Laws, China, Agents, GPUs, AGI", A distinct phase between pre-training (raw knowledge) and post-training (fine-tuning) focused on specialized data ingestion. It involves training on high-quality, long-context data or specific reasoning traces to prepare the model…
