Davuth

Guys! Stacktrace is back (again)!

Oh my! It's true!! 😱

Wait what??? Seriously??!! No way!

I'm using Claude Code. I'm using mostly Sonnet, but Opus is probably not that much better.

And no, it doesn't handle SwiftUI code that well. The way it structures view is not good enough. It generates view code that seems to work on first glance, but it's just not good enough.

No AI model is good with SwiftUI.

Gotta code everything by hand like in the 90s. 😖

I feel that... trying to create an MVP in a week and release it to the App Store is not a good way of building an app.

Not even a good way to actually test the MVP.

Saw on the internet that I'm supposed to "warm up" US TikTok account and stuff to do marketing.

US SIM card... VPN... proxy... new phone... oh man!

This is way too complicated!

My developer brain cannot comprehend this.

Claude is down.

Me: *Google search*
"How to create SwiftUI View?"

Maybe I'll give up half way. Maybe I'm going to hate it. Or maybe it won't make any differences at all . I don't know. Let's see.

Ok. I'm going for it. I'll take 1 month to do this experiment.

My plan:
- Look for a simple app idea
- Find an angle to differentiate it
- Lean heavily on marketing & distribution

That means I'm going to use TikTok for the first time. And learn to make some ads.

Very good perspective on the job.

Maybe I'm just stuck in my own head.

Best of luck!

Doesn't sound fun at all.

Of course, if I could choose, I would just choose to build only the things I enjoy using. But seeing people make money with just making "another habit tracker app" is quite tempting.

Or maybe it only looks easy but it's not that easy? I don't know. I'm torn.

What do you mean by "your heart still has to be in it"?

What if I just... follow this app. Take the MVP features of this exact app. Tweak a few things to make it my own. Then just publish?

But I'm hesitated as I feel that I'm charging users for a less optimal solution. But you kinda make a good point though. Hmm...

You make a really good point.

1. Not all apps need great solution. Some apps (like affirmation, habit tracker) are pretty straight forward.

2. Some users don't pay for great apps. They pay for apps they see. So if ASO god gives the blessing, then it's all that matters.

Is it considered "bad" to build apps for the sake of money?

Not something I'm going to use. Not even interested in the idea. Just pick something purely because it has high demands in the app store.

Of course, no scammy stuff whatsoever. No dark pattern on paywall or anything like that.

I tried this message "Try for $0.00", and my app got rejected ❌

I just want to follow Duolingo app. Apparently, Apple doesn't think this is clear enough for the user.

Time to revert back!

My app is still stuck waiting for review for 3 days now.

Not sure what's going on over there at Apple, but the review process is taking longer and longer these days.

Yep, that's essentially it. Just extract the English (for reference), and the targeted language to work on. Remove all unnecessary keys. Use LLM to translate the yaml and then import back into the string catalog. The file is small and easy to work with.

In GitHub but private. 😅

Too rough right now. Just make it enough for my own use case. Everything is just stuff together to do the conversion.

Another tip is to use Xcode to generate comments for all the string. Then use LLM to go through every comment and improve them if necessary. It helps provide more context for LLM to do a better translation.

P.S. yeah, i'm also waiting for that selfie 😂

I'm using Claude Code for translation. But there are many gotchas along the way. The primary one is that the string catalog is getting so big that the LLM has a hard time working with it. So I vibe coded an internal tool to help with that.

bsky.app/profile/davu...

Why add 1 localization when you can add 11? 😅

Honestly... I underestimated how difficult it would be. I'm so glad it's finally done!

... and now I'm ready for my app to go from 2 downloads per day to 2000, and I'll be drowning in all the profit I'm about to make.

Make sense...

I'm currently using Claude Code (CLI). And it can build and detects error if something is wrong. I don't know how it does it, but it does work. Which is why I wonder what this new MCP will solve.

Do you use AI to help with coding at all these days?

Convenience how? Like... if I install the new Xcode, what can I do that I can't do right now?

And what's your thought?

Sorry if I ask the most dumb question.

I know that this whole agentic coding thing for Xcode is a big deal but I can't wrap my head around why.

I'm already using Claude CLI to code. It seems to be working fine. But now what? What is it now that can be done that can't be done prior to this?