Episode Details

The era of brute-force AI training is dead, replaced by a new frontier where "time to think" is the ultimate scaling law. We tear down the massive architectural shift from training-dominant to inference-dominant scaling, exploring how Monte Carlo Tree Search and semantic routers are turning mid-sized models into reasoning powerhouses. From the logic-bending performance of DeepSeek-R1 to the rise of "System 2" thinking, discover why the most significant pivot since the transformer is happening right under our feet.