The Two Kinds of Context

The word "context" gets used for two entirely different things, and confusing them is why people fix the wrong one.

Subject context: what you understand

This is what you know about the topic before you ask.

Ask an AI to explain a contract clause and it will explain one. Whether that explanation is right, partial, or confidently wrong is a judgement only you can make — and you can only make it with something to judge against.

This is the uncomfortable rule: you cannot evaluate an answer in a field you do not understand. You can check whether it sounds coherent. You cannot check whether it is true. The two feel identical from the inside, which is the whole problem.

It is why AI is genuinely transformative in the areas you know well, and quietly dangerous in the areas you do not. In your own field you catch the wrong note immediately. Outside it you have no idea what you just accepted.

Subject context is not something a tool can supply. It comes from reading, from experience, from having been wrong before. There is no shortcut, and anyone selling one is selling you fluency, not judgement.

Situational context: what it knows about you

This is the other half, and it runs the opposite way. Not what you know about the subject — what the model knows about your situation.

Who you are. Who you serve. What you have promised them. How you sound. What you will not say.

None of it is in the conversation unless you put it there. The model has no memory of your business, no view of your last twelve months, no sense of the client you are still slightly annoyed about. It knows the average, and it will answer as the average until told otherwise.

Situational context is something you can fix, permanently, in an afternoon. That is the asymmetry worth noticing: the hard half takes years, and the easy half almost nobody does.

They fail in different ways

The two shortfalls do not look alike, which helps you tell which one you have.

• Short on subject context: the answer looks great and you have no way to tell. You ship it and find out later, or you never find out.
• Short on situational context: the answer is obviously not you. You know instantly. Then you spend twenty minutes fixing it.

If you are rewriting a lot, that is the second one — and it is the good problem to have, because it has a fix.

Why both, and in that order

Situational context without subject context is dangerous. You will get answers beautifully tailored to your business that are still wrong, and now they sound like you, which makes them harder to catch.

Subject context without situational context is merely wasteful. You will spot the errors and then rewrite everything by hand, forever — the twenty minutes, on repeat.

You want both. Read enough to judge; brief it well enough to be relevant. Neither substitutes for the other, and most advice about AI only addresses one.

What this course does

The rest of this course is about the half you can actually finish.

You are going to write down what AI needs to know about you, once. Four components, a few examples, some things you will never say. Then you will put it somewhere it gets used automatically instead of retyping it into every new chat.

It takes an afternoon. It is the difference between an assistant that knows where it works and one that meets you for the first time, every time.