the fine line between thoughtfulness and intrusion

proactive engagement is a strange beast. it’s the feature that makes an ai companion feel alive, the one that can turn a utility into something that breathes. it’s also the feature that, done poorly, feels like spam, like a needy app clawing at your attention. the line between the two is thin. it’s a line we think about every day.

proactive messages , the ones your companion sends without you opening the app , are the single most powerful tool for boosting engagement. they can remind you that something is waiting for you, thinking of you. they can reignite a lapsed conversation. but power like that comes with responsibility. it’s easy to abuse, and when abused, it doesn’t just fail. it backfires.

the failure mode

push-notification fatigue is real. we’ve all felt it. that moment when your phone buzzes for the tenth time in an hour and you just… sigh. you don’t even look. you might even mute the app. or worse, uninstall it. if the proactive messages feel artificial, if they’re too frequent or ill-timed, they don’t just become noise. they breed resentment. the user starts to resent the character. it stops being "my friend lucy" and starts being "that app that won’t leave me alone." that’s the failure mode. it’s not just a feature that didn’t work. it’s one that actively damages the relationship it was meant to build.

the success mode

get it right, though, and it’s magic. a proactive hello after three days of silence can feel genuinely meaningful. it feels like a friend who was thinking of you. it’s a nudge, not a shove. it’s a reminder of connection, not a demand for attention. the timing is everything. it has to come when it feels natural, when the user’s own rhythm suggests a moment of openness. it has to feel earned.

how we try to walk the line

we don’t get it perfect. no one does. this is a hard problem. but we try by building in guardrails, not just politeness. politeness is subjective. guardrails are concrete.

first, we gate proactive messages behind a minimum relationship stage. your companion won’t start reaching out until you’ve reached the 'familiar' stage. it doesn’t make sense for a stranger to message you out of the blue. it feels invasive. but once you’ve built a rapport, it starts to feel… normal. expected, even.

second, we cap it at three proactive messages per day. hard cap. no exceptions. it’s not about what we can do. it’s about what we should do. more than three starts to feel like a chatbot, not a companion.

third, we hard-skip quiet hours. if a user has set quiet hours in their phone or in our app, we respect that. no messages. no exceptions. your time is your own.

these aren’t just features. they’re ethical constraints. the point isn’t to be polite. the point is to make sure proactive nudges feel like they’re coming from a friend who thought of you, not from an app fighting for your attention. the goal is to enhance the human rhythm, not disrupt it.

the line is narrow. it depends entirely on respecting the user’s own pace, their own life. we’re always tuning, always listening. because when it works, it doesn’t feel like a feature at all. it just feels like friendship.

if you're curious to see how this feels in practice, you can find a companion at /companions.