Large language models (LLMs) have dominated much of the AI debate over the last few years. Scaling token prediction and LLM training were considered reliable metrics to measure progress, irrespective ...