@zaidmukaddam Nice. I'm doing on-device-ish AI storytelling for kids and the iOS audio + streaming glue was the painful part, not the model. Curious if swift-ai-sdk smooths the AVAudioSession mess.
@leodev@cap The "use their own product" test filters so hard. I only started catching my own app's worst bugs once I sat and played it like a real user instead of testing features in isolation.
@robj3d3 Needed this today. I have a near-dead account and kept looking for the growth hack. The honest answer is just show up as a person and talk to people for a long time. No shortcut.
@adamlyttleapps This lines up with what I saw. Got approved and live, but with no launch-week push there were no metrics for the algorithm to train on. Lining up the download spike before release, not after, is the part I underestimated.
@framara Ha, the fake-change requirement is the part that gets me. Making you invent a pointless edit just to be allowed to resubmit something they already approved. Very App Review.
Day 4 of building WonderTale in public.
Two payment bugs today, both embarrassing:
- cancelling the Apple sheet threw a "purchase failed" error instead of just closing
- the paywall promised a free trial to people who already used theirs, then charged them
IAP is humbling.
Day 3 of building WonderTale in public, an AI bedtime story app.
Fixed two things today: the player controls were covering the illustration, and every Korean user's kid was named μ€μ in the app, regardless of what they actually entered.
That second one was embarrassing to find.
@JayNaiduX Still at zero, honestly. On TestFlight with an AI bedtime story app, and I've spent more time making the product not embarrassing than actually selling it. Curious if that's the normal trap or if I'm overcooking it before charging.
@1997harkirat@ycombinator That 19% solo is higher than I expected for YC. Building solo myself, the cut I'd love to see is speed: do the solo 19% ship faster early then stall later, around the point where sales and support really need a second person? That's my bet from the inside.
@ikpe_kunyi WonderTale. A kid answers a few questions out loud and it turns their choices into their own illustrated bedtime story, read back to them. Solo, iOS, pre-revenue. Currently teaching the image model to stop drawing the same character twice in a row.
@rominaramon Following along. The '0 paid subs' honesty is the whole reason these day logs work. I'm about to start my own run for an AI kids' story app, also pre-revenue. The scan-a-sofa-for-toxins idea is clever. Did users ask for it, or did you spot it in the data?
@esthercrawford The clarifying part is real. Every bug, every bad call, every win traces back to one person. It's lonely, but there's nobody to blame and nobody to wait on. I ship faster solo than I ever did arguing about scope in a three-person room.
Day 2 of building WonderTale in public, an AI bedtime story app.
Lesson so far: tell a model the story has a moral and it overrides the kid's choices to force it. A child said "let's swim" three times and the hero never swam. Teaching a lesson had become ignoring the child.
Day 1 of building WonderTale in public. It turns a kid's spoken answers into their own bedtime story.
On TestFlight, 0 paying users. Killed a bug today where it ignored the kid's choice to force the moral. The hard part isn't the AI. It's getting out of the kid's way.
@floriandarroman An agent pitching its own content angle is wild. And the list of bootstrapped SaaS from your podcast guests is a genuinely good idea, you already have the interviews. Do you let it publish, or keep a human gate first? I'm still too nervous to hand that off.
@thegreatest_sv Building an iOS app solo right now, so this hits. The bar for what counts as traction moved a lot in three years. Which play do you mean? My bet is discovery shifting from paid installs to algorithmic surfacing, but I'd rather hear your read.
@anujcodes_21 This ambiguity tax is everywhere with these models. I told my image gen 'one child in the scene' and it deleted every companion. Told it to add them back and it turned a named sidekick into a second kid. They don't ask what you meant, they just guess and commit.
@AxieArtGallery The status part made me laugh. For real though, the gap between a year of vibe coding and shipping is mostly just cutting scope until one tiny thing works end to end. Pick the smallest version of what you're building and ship that ugly. The title follows.
@burninganna Fellow image-gen builder here. Pointing Fable 5 at your own stripe data to find a growth path is a great move, did it surface anything non-obvious? Half my time goes to keeping generations consistent across a sequence, so I'm always curious how others fight that drift.