The route: Classical Machine Learning, then Neural Networks, then Tokens and Embeddings, then Transformers, then Training and Tuning, then Inference and Serving, then Prompting, then Evaluation, then Search and Retrieval, then Agents, then Production AI.