Few-Shot and Chain-of-Thought Patterns

The RTFC framework tells the AI what you want. The two patterns in this lesson go further: they show the AI what good looks like, and they slow it down enough to think. Layer either one on top of an RTFC prompt and quality jumps; layer both and you are operating well above what most people ever get out of these tools.

Few-Shot Prompting: Show, Don't Tell

Few-shot means giving the AI one to three examples of what you want before asking for the real thing. The "shots" are the examples. Zero-shot — what most people do — is describing the task with no examples at all and hoping the description is enough.

Instead of describing the format in words, you demonstrate it. The model is a world-class pattern-matcher, so when it sees two captions in your voice, it infers the length, rhythm, and tone far more accurately than any adjective could convey. You stop explaining and start showing.

This works especially well for tone, length, and structural consistency — exactly the things that are painful to describe but easy to demonstrate. Two or three good examples is the sweet spot.

Chain-of-Thought: Make AI Think First

Chain-of-thought prompting tells the AI to reason through a problem before it answers. Left to its own devices, a model often blurts a confident first guess; forcing it to work through the steps produces noticeably better judgment.

You trigger it with a single line: "Think through this step by step before giving your final answer," or "First lay out the factors, then weigh them, then recommend." That instruction alone changes the quality of the response on anything with moving parts.

Use it for complex tasks — research synthesis, strategic decisions, pricing, anything with multiple factors to balance. For simple generation it just adds noise, so save it for the moments where being right matters more than being fast.