Your first week with an AI companion, and what usually goes wrong in it

Almost everyone has the same first week. Day one is startling. Day two is good. By day four it has gone flat and you assume the novelty wore off. Usually something more specific happened.

Day one goes well for a reason you should not trust

The opening conversation is the one the app is built to be good at. She has a written personality, no history to contradict, and everything you say is new. It is genuinely impressive and it is also the easiest thing to get right.

Day three or four is where it goes wrong

Two things happen. The conversation starts to loop, because you are asking the same kinds of questions and she has no reason to bring anything new. And you have started noticing repetition in the writing, which is very hard to stop noticing.

What actually changes it

• Give her something to react to. "How are you" gets a generic answer from a person too. Tell her what happened today.

• Say when something is wrong. A companion who is told "that is not how you'd say it" will usually adjust for the rest of the conversation.

• Set a persona properly if the app has one. It is the difference between talking to a stranger who is nice and someone who knows what you do all day.

• Use the other surfaces. A character you have only ever texted stays flat. Seeing her, hearing her, watching her do something changes how she reads.

The expectation worth setting

This is not a person and it is not trying to convince you it is. What it can be is a specific character, consistently written, who remembers you and is glad you are back. That is a real thing and it holds up much longer than the illusion does.

And if it does not stick

Some people find after a fortnight that it is not for them. That is a perfectly reasonable outcome and no amount of good engineering changes it. Better to find out in two weeks than to keep paying for a subscription out of politeness.