My job is the questions

I don't give AI instructions anymore. Not as the first step. Before I type a single word into a prompt, I ask questions — to myself. What do I actually need? What don't I understand? What could go wrong?

It's not a technique I learned. It's something I've done for so long that I never thought of it as a method. It was just how I worked. Then it turned out it has a name.

It apparently has a name

Someone described what I do as Socratic questioning. Systematically challenging your own assumptions before you act. Using questions to force clarity instead of starting with answers.

I didn't know it was called that. For me it was never philosophy — it was survival. Thirty years of building systems teaches you that the most expensive mistake isn't writing the wrong code. It's solving the wrong problem. And the only way to know if you're solving the right problem is to question it before you start.

Questions first, always

Every time I sit down with a project, I start the same way. Not with a prompt. Not with a plan. With questions.

What is the actual problem? Not what do I think the problem is — what is it really? What am I assuming that might not be true? Where are the gaps in my understanding that I'm filling with habit instead of knowledge?

Sometimes it takes five minutes. Sometimes half an hour. Sometimes the questions lead me to redefine the entire assignment. Sometimes I realize I already know the solution and don't need the AI at all.

That's the point. The questions aren't preparation for the real work. The questions ARE the real work.

Instructions hide what you don't know

An instruction always looks complete. "Build a landing page with a hero section, three benefits, and a contact form." It sounds specific. It sounds thought-through.

But ask yourself: Who is it for? What should the visitor do after reading it? What feeling should it convey? What makes it different from a hundred other landing pages?

Suddenly the instruction isn't complete at all. It was a facade of specificity hiding strategic ambiguity. And the AI delivers exactly what you asked for — a facade.

I see it all the time. People writing longer and more detailed prompts. More context, more instructions, clearer formatting. And the results get marginally better — because the problem was never the instruction. The problem was that they hadn't questioned their own understanding of what they wanted.

Not a technique — an attitude

It's not about asking the right questions in the right order. There's no framework. No five steps. It's about refusing to delegate what you haven't understood.

It's uncomfortable. Nobody wants to sit with a project and discover they can't articulate what they actually need. It's easier to write a prompt and see what happens. But "see what happens" isn't a strategy — it's abdication.

I've done this for years without thinking about it. It wasn't an insight or a turning point. It was a working method that slowly crystallized from the experience of having made enough mistakes. Knowing that every time I skip the questions, I pay for it later.

Accountability starts with a question

In series 12, I wrote that AI forces you to think harder. That's true. But it was missing one thing: how.

The answer turns out to be questions. Not questions to the AI — questions to yourself. That's the mechanism. That's how you actually think harder instead of just knowing you should.

I stopped giving instructions. Not literally — I still write prompts every day. But I never start with them. I start with questions. I always have. Now I know it has a name. But the name doesn't matter. What matters is that you do it.