It’s extremely disappointing and frustrating to see some new service that does not disclose how it’s expected to “keep the lights on” and how that would impact people who sign up.
I need the answers on the site, even if they’re provided here as a reply.
We built it and wanted to ship now and figure that out later. Obviously our goal is to keep the service up and we will. We're not strapped for cash. Ultimately we'll charge some kind of yearly fee for people who have a lot of decks, but have not come up with a monetary amount or what limits would exist. Our assumption was always to grandfather in early accounts (such as those created now) and learn from people's feedback.
Seeing it verbalized that this is frustrating is important, though, so thank you.
We'll add that as context on the site, and think through what a paid plan might look like.
I've always been interested in spaced repetition but never had the patience to learn the "right" way to do it. This looks pretty helpful.
Do you think this AI x repetition concept works better for some types of learning than others?
Owl wasn't built with AI in mind, though. It isn't necessarily an AI product. We use it where it helps (analyzing a paper and creating cards to study that paper is very cool, for example). I think there are even more angles to explore, but we can't claim to have tried them all :-)
I saw you said you didn't like the Anki UX, and I agree that it's a little unfriendly to new users. Your product looks like it's well pitched to new SRS users. But I don't see anything the compels, or even proposes, a need to switch from other solutions right now.
> Application error: a server-side exception has occurred (see the server logs for more information). > Digest: 2072004716
Almost had me prior to sign-up