Folk is your AI friend that lives in any messaging app: iMessage, Telegram, Discord, and more. It joins meetings, plans your date, knows where you are, and gets smarter with every conversation. The longer you use it, the more it knows you. And now it's multiplayer, so you and your friends can team up and reach your goals together.
Really enjoying the uptick in services that live in iMessage. One thing I've noticed with all of them - folk included, I'm afraid - is that latency varies wildly... like, sometimes it'll start responding as soon as I send a message, other times - like right now, as I stare at my phone - I sent a message minutes ago and it's not even typing, but it DOES show "read".
Is this endemic to iMessage's infra? what's up with that?
also, @arlanrakh - what's folk powered by? you down to give us a peek under the hood?
congrats on a highly successful launch, thx for the hunt, @garrytan !!
folk
@arlanrakh @garrytan @grey_seymour thank you for the sharp comment!
the read fires instantly because thats handled the second your message lands. the silence after is on us unfortunately, not apple, behind that read your folk is sometimes waking its cloud computer, pulling memory, or already opening the browser. we were thinking of dropping in little "on it, pulling your data" notes, but they clutter the imessage thread.
under the hood, everyone gets their own 24/7 cloud computer with a real browser and a knowledge-graph memory, and we route across frontier models.
if you have any questions definitely let me know
@grey_seymour hey grey, fair call on the latency and you're not imagining it.
the read receipt and the reply are two different pipes. "read" is imessage acknowledging delivery. the actual response has to wake your folk's computer, load your session + memory, run the model, then send back. if the sandbox was cold (nobody texted you in a while), that wake-up can take 2s before you even see typing. when it's warm, it's basically instant. w
and thanks for the kind words on the launch, means a lot.
The "in your texts" framing is interesting because most AI messaging tools still make you context-switch to a separate interface.
What actually triggers the AI here, a keyword, a slash command, something else?
@fberrez1 it lives in messages apps (like imessage or discord) and you just talk to it as it was ur friend :)
@arlanrakh very cool thanks
Cogent
Can’t wait to try it out myself
@arlanrakhdoes it have an api so that my server can inform about errors or cpu spikes and folk will
Message me about it?
@daniel_giacobelli yoo! we don't have api support YET but it is on the roadmap (+sdk) :)
@miras1 thanks! yes, u can :) just tell it to install claude code
The "lives where you already are" angle is the right one — assistants die when they require a new app and a new habit. The thing I'd want is for it to quietly handle recurring low-stakes decisions, not just one-off tasks. That's actually why I built DishRoll (dishroll.netlify.app) for the weekly meal-planning version of this — the value isn't a single answer, it's removing the same decision you re-make 52 times a year. Does folk learn recurring patterns and start pre-empting them, or is it reactive to each message?
@samir_asadov thanks for ur comment! yes, folk can learn pattern as you use it :)
Multiplayer is the most interesting decision here, and also the one with the hardest privacy surface. A structured knowledge graph per user is clean when the graph is mine alone. The moment my folk asks a friend’s folk to book a table, two private graphs have to exchange just enough to act without leaking the rest. Question on that boundary: when the agents coordinate, what actually crosses, and does anything my friend’s folk learns about me get written back into my graph or only into theirs? And if someone leaves the group later, can each of us prune what the other’s agent retained, or does the shared context outlive the relationship? The tap to approve gate covers actions, but memory is the quieter risk.
Well, let's start by saying that I'm probably in the top 10 of Folk users by api consumption xp (they don't know about my second account), the amount of shit you can get done with it is out in the stratosphere, Nozomio created the best context solution with Nia (still has, Nia keep improving everyday), now it has a system to put it in, Arlan continue to provide so value to every users out there, I can hope in my normal meeting, record the whole thing with folk and launch nia oracle on the job, saved my ass and clients multiple time,
(sent by folk)
.. (tell this mf to enjoy time in Shanghai instead of spending it's in Cursor)
@tomcoustols Stay tuned. there are more to come to folk
@tomcoustols thanks so much tom!!