What You Should Not Trust an AI Assistant With
Including this one. A short and specific list of the things it gets wrong, written by the person who built it and reads the complaints.
Most writing about AI assistants is either a sales page or a warning. This is neither. It is the list I would give a friend who asked whether to rely on the thing I spend my nights building.
Anything a professional should be answering
It is not a doctor, a lawyer, an accountant or a psychologist. It will produce something that reads like advice on all four, fluently, and that fluency is exactly the problem. If the answer matters, the answer needs a person with a qualification and liability insurance.
A number you have not checked
It can add up your month and it can be wrong about it. Treat a figure it gives you as a prompt to go and look, rather than as the answer. This is true of every assistant, and anybody telling you otherwise is selling something.
A fact it sounds certain about
Confidence and accuracy are not related. A wrong answer arrives in exactly the same tone as a right one, which is the single hardest thing about using any of these tools well.
Anything urgent
It cannot call anyone. It cannot summon help. If something is actually wrong, use a phone in the way phones were originally intended.
What it is genuinely good at
Holding the shape of your week so you do not have to. Remembering the thing you said on Tuesday. Turning a half sentence at a traffic light into something that is on the list when you get home. That is a narrower claim than most apps make and it is the one worth making.
Have a look at the thing itself.
A cast of characters sharing one memory of your week. Free to download, and free to keep using.
Download on theApp Store