I have one framework I use for every prompt I write. It’s how I make sure the AI has enough context to actually do the job. It’s extremely simple.
Before I send a prompt, I look at everything I’m giving the AI: the prompt, the context, files, everything it might need. Then I ask myself one question.
If a brilliant intern (let’s say a kid who just graduated from Harvard) walked into my office on their first day, knew nothing about my company, knew nothing about the work, and I handed them a piece of paper with exactly what I just gave the AI and asked them to do the task, would they be able to figure it out and do it as well as I would?
If yes, the AI probably will. If no, the AI probably won’t.
That’s it.
I know it sounds too simple to matter, but in my experience, 99% of prompting failures come from not giving the AI everything it needs to do the job. Especially as the models get smarter, they don’t need tricks or clever wording. They just need all the stuff a human would need to do the same job. Give them that, and you’re good to go.
One example. If you paste a prompt like “Plan me a trip to Japan,” the intern wouldn’t know when you’re going, what you want to spend, who’s coming, or what you actually like. But if you add your dates, your budget, “two kids, 6 and 10,” and “we hate tourist traps,” the intern could plan you a real trip. So could the AI.
So that’s the rule. Don’t send the prompt until it passes the intern test.
Try it on the very next prompt you send.
If you want an easy way to make sure your prompts pass the intern test every time, I built a tool into Something Big Pro that does this for you: paste in any prompt and it rewrites it to pass. I added a bunch more to Pro too. You can see it all below.
Hit reply and tell me if it changes anything.
— Matt
What should I write about next? Vote or suggest a topic →