ben-gy

It’s possible, but it would look unnatural to the human eye - you could introduce variable motor speeds to try and mimic bird behaviour when flying in different directions and approaches but from my rabbit holing on the topic I’ve found effort required to get something similar but still not perfect would likely outweigh the effort required to just build it properly.

I have deep appreciation for well crafted vanilla front-end and for the value micro interactions bring to the table, but I really struggle to reconcile the value of this now that we have LLMs that do a phenomenal job at solving these UI/UX problems.

Tailwind UI is a perfect example of this where when I used to use Tailwind UI religiously and paid for it, I can just tell Claude Code “make it much prettier and bring it to life with really well designed micro interactions to create an excellent and enjoyable end-user experience” and it will get me 90% there which is easily more than good enough.

I really hope I’m wrong and that there is still a market for hand crafted components like this but honestly I think the free components exposed here plus the images of the paid components is enough for any LLM to pick this up and run with it likely without the LLM user even knowing that’s what has happened and therefore I can’t see a path to sustainable ARR for you guys.

This is interesting - I’ve been messing around with AI generated games - I get about 50% out-of-the box ready to play intuitively now but it took a lot of refinement of the prompt to get there - I’ll tell my ai to have a go at this one and see what it produces (will credit op for the inspiration).

I would say this is an important part of the product-market-fit discovery process. Sure, your comment holds true for well established products with a loyal customer base, but for an earlier stage product this rapid iteration process is quite important, especially from an onboarding and ease-of-use perspective.

This feels like force majeur from a contract perspective…

Opus 4.5 got released ~3 months ago - Claude Code started using it automatically (for me anyway) - I also tried iOS prior to that and had a similar experience to you

Max

Very few - the only manual things I do are; - clicking the distribute button to push the bundle to the App Store - filling in the compliance survey and App Store listing content - linking some components together e.g. for creating a VPN installer and tunnel i had to click some things in the Xcode UI I automate as much as possible; -“create 12 app icons for this in SVG and present them to me in a HTML page so I can choose one and then use that for the app icon and splash screen” - “create a demo mode toggle in settings and populate the app with fake data and then open up simulators for the correct image dimensions for the App Store listing so I can crate screenshots” - sometimes it tell me I have to other things like set up the entitlements to which is say “no - you do it and don’t forget to fill in the description that gets shown to the user so the feature actually works” I knew very little about Swift or Xcode profile to this and TBH I still don’t know that much about it, but I’m experienced enough to know when I’m being fed something that doesn’t look or feel right programmatically or architecturally.

I second this article - I built twelve iOS/Mac apps in two weeks with Opus 4.5 - four of them are already in the App Store - I’m a Rails Engineer and never had the time to learn Swift but man does Opus 4.5 make that not even matter - it even handles entitlements, logo & splash screen generation, refactors to remove dead code, edge case assent and hardening, Multiplatform app design, and more - I’m yet to run into a use case it can’t handle for most general use cases - that said, I have found some common mistakes it makes (by common I mean almost every time); puts iOS line list line items in buttons making them blue when they should not be, doesn’t set defaults for new data structure variables which crashes the app when changing the data structure after the fact, design consistent after the first shot (minor things like white background instead of grey background like all the other screens already, etc) - the one thing that i know it cant do well (and no other model that I know of can do this well either) is ASTM bi-directional communications (we work with pathology analysers that use this 1995 frame-based communication standard), even when you load it up with the spec and supporting docs - I suspect this is due to a dirty of available codebases that tackle this problem due to its niche and generally proprietary nature…