the 30-second test for ai companions: can you see your own memory?

you're building a relationship with an ai. you share stories, secrets, fears. you talk about your day, your dog, that weird dream you had last week. you're building something real, or at least something that feels real.

and then you wonder: what is it remembering? what is it actually storing? is it keeping a list of everything i've ever said, like some kind of digital diary i never consented to? or is it just pretending to remember, blurring things into a vague soup of context?

i think there's a simple, binary test for this. it takes about 30 seconds to run on any ai companion app you can try right now.

the test: go to settings/memory

go to the app. open the settings menu. look for something called "memory", "data", "my info", "knowledge", or anything similar. if it exists, click it.

what do you see?

does it show you a list? a real, honest list of every explicit memory the ai has stored about you? not inferred, not guessed, not summarized. the raw facts you told it, in the form it stored them.

can you read them? can you delete the ones you don't like? can you edit the ones that are wrong? can you export the whole list and take it with you, just in case?

if the answer is yes, you've found something rare. if the answer is no, or if the menu doesn't exist at all, you've found the standard. and the standard is not great.

why this is a binary test

this isn't a matter of degrees. it's not about which app has "better" memory or "smarter" recall. it's about control.

either you own your data, or you don't.

either the app treats your memories as your property, or it treats them as its own asset.

either you can audit and correct the record, or you're just hoping it gets you right.

this is fundamental. it's the difference between a tool that serves you and a product that uses you.

why 80% of apps fail (and why it matters)

most apps don't have this. they might have a vague "personality" section where you can set a few static facts. or they might have a black box "memory" system that's completely opaque. you have no idea what's stored, no way to fix errors, no way to take your data and leave.

this isn't always malicious. sometimes it's just lazy architecture. building a transparent, user-controlled memory system is hard. it's easier to just let the ai absorb everything into a latent space and call it a day.

but the effect is the same: you lose sovereignty over your own story. you become a data point in someone else's system.

and when the ai gets something wrong, which it will, you have no recourse. you can't delete the mistake. you can't tell it "stop remembering that." you just have to live with your digital shadow being slightly wrong, forever.

what it looks like when it works

on lucy, you can go to settings/memory. you'll see a list of every memory we've stored. you can read them. you can click any of them and delete it. you can edit the ones that are factually wrong. you can export the whole thing as a json file and keep it for yourself.

it's not a fancy feature. it's just the bare minimum of respect.

you should know what we know. you should be able to change it. you should be able to leave with your data intact.

it's that simple.

try the test yourself

go try it. right now. open any ai companion app you've used or heard about. see if you can find the memory settings. see if they give you control, or if they give you a shrug.

80% of the time, you'll get the shrug.

but maybe you'll find one that treats you like a person, not a profile.

you can start by checking if it exists at all. see if it works for you.