Discover How to Solve Any Problem From Scratch Using AI

Reed Hastings paid a $40 late fee on a VHS copy of Apollo 13. According to the story he told most often (Netflix's own co-founder has said it's a bit exaggerated, but Hastings insists it happened), that fee is what made him ask why video rental had to work the way it did. The obvious answer would've been "because that's how video stores work." He didn't accept the obvious answer, and Netflix was born out of refusing it.

Steve Jobs did something similar, only with a phone. In 2007, every serious device had a physical keyboard. The BlackBerry, the best-selling phone at the time, lived off that fact. Jobs looked at the keyboard and asked whether it was actually necessary, or just a part every manufacturer copied from the last one without questioning it. He removed the keyboard. You already know how the rest turned out.

Both of them asked the exact same question, about completely different problems. And that question turned into a prompt you can copy by the end of this piece.

The thought that crosses your mind and dies before becoming action

You've already had this thought. Not as some big insight, but as a thirty-second complaint that passes and disappears. "Why does my phone plan cost $120 when I barely use half the data?" "Why does support at my company need to respond within 24 hours, who decided that?" "Why does this Monday meeting still exist, when nobody even remembers why anymore?"

The thought shows up in the shower, in traffic, two minutes before falling asleep. And it dies right there, because following it through takes work: you'd have to list every piece of that rule, test each one, and probably give up on the second piece because nobody has the patience to do that alone after a full day of work.

It's exactly that thought, the one that usually dies in thirty seconds, that the prompt below turns into a real answer. This has a name: first-principles thinking.

Why an AI can keep asking why until the end

You've probably already tried to question something like this and given up halfway. Ask your phone plan why it costs what it costs, someone from support says "it's the standard plan," and you hang up without pushing further, because nobody has the patience for that on a ten-minute call. An AI doesn't get tired. It takes each piece of a practice, whether it's a business one or a personal one, and tests them one by one, with no rush to end the conversation.

I have a problem I currently solve (or plan to solve) the conventional way, the way everyone does it. I'm going to describe the problem and how it's normally solved. I want you to:

1. List every assumption embedded in this conventional way of solving the problem, even the ones that seem too obvious to question.
2. For each assumption, classify it as a real constraint (physical, legal, financial, something that wouldn't change no matter who solved the problem) or a copied convention (nobody ever tested whether it was actually necessary).
3. Using only the real constraints, build a solution path that's different from the conventional one, ignoring every convention you identified.

Problem: [describe the problem, personal or professional]
How it's solved today, conventionally: [describe the standard practice you follow or observe in your field]

Run this on a pretty everyday situation, like a $120-a-month phone plan you've never reviewed, and the output looks something like this:

Assumption: I need unlimited data because I use my phone all day.
Classification: copied convention. Real usage history (if you have access to it in the carrier's app) usually shows consumption well below the contracted limit.

Assumption: a plan with more data is always the better deal.
Classification: copied convention, pushed by the carrier, not based on your actual need.

Real constraint identified: you need stable signal in the places you actually use your phone (home, work), and enough data for the apps you actually use.

Alternative path: a plan with half the data, checked against your real consumption over the last three months, cuts the cost by almost half without losing anything you were actually using.

In a business, the same reasoning takes down things like "support needs to respond in real time, 24 hours a day, because that's what every competitor does," when in practice only a fraction of tickets actually block product use and need an immediate response.

Why this mapping works

The prompt works because it forces the AI to take the entire practice apart, assumption by assumption, instead of giving a loose opinion about it. Nobody does that work of holding several assumptions in mind at once during a ten-minute call center call or a thirty-minute meeting. An AI does, because it's in no hurry to hang up.

The more detail you give about how the thing is solved today, the better the result, because it can only question the assumption you made visible to it.

Reed Hastings had a late tape. Steve Jobs had a keyboard nobody questioned. You have something that's always just been that way because you never asked again. The difference between their cases and yours isn't genius. It's that they stopped to ask the question, and most people never do.