Ten-Cent Inference and the New AI Distillation War

We break down how a 1M-context Mixture-of-Experts model can deliver frontier-level coding performance at near-baseline pricing, and why its local, open-weights design is changing what developers can run on their own hardware. The discussion also digs into the brewing distillation wars, including White House accusations, the legal gray area around model training data, and the growing tension between open models and regulatory pressure.