Reusable prompts and learning by example
Prompt templates scale
A prompt template is a reusable prompt with placeholder variables. For example, a skills platform might insert a user's job title, level, and preferred language into the same template, personalising the output for each person without writing a new prompt every time.
Act as X
One of the most popular template patterns is 'act as [role]'. Open-source prompt repositories like the Awesome Prompts collection on GitHub contain thousands of these. They work because assigning a role narrows the model's behaviour to a relevant subset of its knowledge.
Zero-shot vs few-shot
Zero-shot prompting gives the model a task with no examples. Few-shot prompting includes examples of correct input-output pairs. Few-shot is especially powerful when your definitions differ from the general case — like telling the model that 'it is okay but I expected more' counts as negative, not neutral.
Few-shot as a living dataset
Some AI startups treat their few-shot examples like a growing dataset. When a user flags an error, a human labels the correct answer and adds it as a new example in the prompt. This is faster than fine-tuning because you never touch the model's weights — you just update text.
If you had to classify 10,000 customer reviews using your own definitions of positive and negative, how would you teach the model your rules?