Fixed
Write a tweet about remote work.
One instruction, many inputs.
So far, every prompt has been fixed text. But in real systems, prompts are usually templates. A template contains placeholders that are filled at runtime. Example: ``` Explain {{topic}} to a beginner in 3 bullet points. ``` The instruction stays the same. The input changes. At runtime: - topic = batteries - topic = solar panels - topic = HTTP The same prompt structure works for different inputs without rewriting the instruction itself. This separation between instructions and runtime data is one of the foundations of instructional engineering. You are building reusable behavioral systems, not single-use prompts. Different platforms use different placeholder syntax: - {{topic}} - {topic} - $topic The exact syntax matters less than the concept: separate the instruction from the data.
Write a tweet about remote work.
Write a tweet about {{topic}}.
The template version scales across many inputs.
Templates make prompts reusable by separating instructions from runtime data.
Start with a template prompt and an Input node. Run it, see the output, then improve the template. Try changing the Input value to see the same instruction work for different data.