Rather than learning the "techniques," it might be more appropriate for newcomers to understand the "principles" first (as changes don't happen that quickly). For example, understanding the working principles of LLM, time series analysis, what context is, what distillation is, what constraints are, and what the consumed tokens actually are… You'll understand why illusions are unavoidable? You'll understand why some people infer AGI from token consumption patterns? You'll understand why specialized agents are currently used, rather than a single agent performing all the work, etc. More than these, the greatest alpha for ordinary people is actually leveraging AI to quickly understand a cutting-edge/specialized field.