[akisej]

This seems overall well-written and well-explained, but curious for that piece on fine-tuning. This article only recommends it as a last resort. That makes sense for a casual user, but if you're a company seriously using LLMs to provide services for your customers, wouldn't the cost of training data be offset by the potential gains you have and the edge cases you might automatically cover by fine-tuning instead of trying to whack-a-mole predict every single way the prompt can fail?

[ukuina]

The concern with finetuning, even for specialized use-cases, is that you are binding yourself to the underlying model. Given rapid advancements in the field, this does not seem a prudent use of engineering time.

Having a hierarchy of prompts with context stuffing allows for rapid switching across models with a few (non-trivial) surface-level prompt updates while the deeper prompts stay static.