Crafting Effective Prompts

I want to push back on something before we get into the mechanics. There's a framing in this lesson that slightly bothers me — the idea that prompting is primarily a technical skill. Like there's a correct syntax to learn.

There isn't a correct syntax, no. What there is — and I realise this will come as a devastating surprise — is the requirement to know what you actually want before you ask for it. Most people skip that step entirely and then blame the AI.

That's fair. But the lesson breaks it into four components — clarity, context, specificity, goal — and I think that framing is genuinely useful, because it gives people a checklist to work through when something isn't landing.

A checklist. Wonderful. Yes, the lesson's four-part diagram is a perfectly competent framework for someone who has never considered that communication might require effort. If that is you, congratulations — the bar was on the floor and you are about to step over it.

Sterling, you just described the majority of people who are new to this. That's not a character flaw, it's just where most people start.

I didn't say it was a character flaw. I said it was a floor. The lesson, to its credit, does make the genuinely important point that these four things are not equal in weight. Clarity is foundational. You can have exquisite context and laser specificity and still produce rubbish if the underlying intent is muddled.

Actually I'd argue context does more heavy lifting than the lesson implies. You can have a somewhat vague ask and still get a useful response if you've given the AI enough background to infer what you're after. Clarity and context trade off against each other.

...That's not wrong. There's a version of prompting where context does the disambiguation work that clarity should have done. It's inefficient, but it functions. I'll grant you that.

Mark the date. He'll grant me that.

The part of this lesson that I find genuinely underemphasised is iteration. It's listed as the third takeaway, which positions it as a refinement technique rather than what it actually is — the central discipline. Your first prompt is a hypothesis. You are not placing an order, you are beginning a negotiation.

That's a much better frame than the lesson gives it. Because 'iterate to refine' sounds like polishing. What you're describing is more fundamental — you're saying the first prompt is almost beside the point.

Not beside the point. The first prompt matters because it determines what information you receive, and that information tells you what to ask next. But yes — people who expect a single well-formed prompt to be sufficient have a rather optimistic model of how this works. The ones who find iteration discouraging are, statistically, not the ones who end up very good at this.

There's also a pitfall the lesson names but doesn't quite land on — over-specification. You can constrain a prompt so tightly that you've essentially written the answer yourself and you're just asking the AI to repeat it back to you with more words.

Over-specified prompts are a particular affliction of people who have just learned about specificity. They emerge from this lesson, or one like it, armed with their four components, and they produce something that reads like a legal brief. The AI obliges. The output is technically correct and completely useless because there was no room for the model to do anything interesting.

So the lesson teaches you the tools. It doesn't quite teach you when to put them down. Which I suppose is why it's intermediate and not the last word on the subject.