Drawing a map of what I do not know, with AI

The scariest part of starting something new is not what you do not know. It is not knowing what you do not know. The first you can ask about. The second you cannot, because you do not know where to point.

Socrates spent hours of dialogue leading people to exactly that spot, the moment they admit "I do not know this", which the Greeks called aporia. He saw it as the starting point of learning. I get to the same spot with Claude in about an hour. This post is what I ask in that hour, and one map I drew that way.

The use of a dead end

Real questions only appear once you know you do not know. Before that, questions carry a trace of pretending, and the answers match the pretence.

The trouble is you cannot reach that spot alone, because you cannot point at your own blind spot. Someone outside has to ask. That someone is now always available.

Five questions that draw the map

These are the questions I ask in the first hour of entering a new field. They are not about the subject. They are about the shape of my ignorance.

One. What do beginners in this field usually miss? Other people's holes are probably mine.

Two. What do experts take for granted and never explain? What is never explained is not in the books either. You only get it by asking.

Three. Is there a wrong assumption in what I just said? I write a paragraph about my situation first, then ask. This one catches the most.

Four. What is the most common way this goes wrong, and when does it show? That is the risk map. Knowing when it shows lets me check early.

Five. What have I not asked yet that I should? Same as the last of the seven questions. This one fills the blank areas of the map.

One of my maps

This is the map I drew with those five the first time I opened my company's corporate tax return, after ten years of leaving it entirely to an accounting firm. Three columns.

What I knew. If you hand it to the firm, the return gets filed; money goes out by direct debit every month; one more invoice arrives at year end.

What I thought I knew. I assumed the money paid to the firm was one thing. Question three, is there a wrong assumption in what I just said, caught it. It was two things. The monthly fee is for bookkeeping and the VAT and withholding filings; the year-end fee is for closing the books, the tax adjustments and the corporate return itself. For ten years I had held them as one.

What I did not know. Which box on the return I should actually look at; that some companies are allowed to file without an accountant at all; that last year's return can be downloaded from the tax office portal without asking the firm. These came from questions one and two, the things experts take for granted and never explain. This column was the longest, and it generated the next round of questions.

After drawing the map, what I did was not "study tax". I pulled out last year's return and looked at that one box. Then I took each item in the last column and asked why is that so, one at a time. I could ask because I now knew where to point. The reason no question had come out of me in ten years was not that tax is hard. It was that this map did not exist.

Redraw it monthly

Boundaries move. Something I did not know last month moves into the known column, and behind it a new unknown becomes visible. So I ask the same five questions once a month. If the answers match last month, that field can rest. If they differ, the map gets corrected.

Socrates brought people to a dead end and started the conversation there. I find the location of that dead end in an hour. After that it works as it did in his day: you ask from there.