Claude Code Can Now Build and Test iOS Apps in Apple's Simulator

[cjbryce]

Harsh, and possibly true, but then there's folks like me with good ideas but no computer science degree who have been empowered with AI. I learned some python from evening classes (my day job is in life sciences) and could only write simple data transformation scripts that would take me weeks to debug.

With AI - I'm writing creating genuinely useful programs I could never have dreamed of.

Let's not be totally down on AI!

Agreed.

I was a developer for about 20 years until I moved away from that. Now I run businesses and AI has allowed me to develop and deploy new systems and features in days, which would have taken me months to hand code.

Whilst there maybe lots of slop out there, professional developers shouldn’t underestimate the power and advantage AI can give them as it has done me.

 

[tkermit]

All the more reason for avoiding iOS apps. And more generally, why I don't pay for software anymore.

 

[Luftkopf]

Harsh, and possibly true, but then there's folks like me with good ideas but no computer science degree who have been empowered with AI. I learned some python from evening classes (my day job is in life sciences) and could only write simple data transformation scripts that would take me weeks to debug.

With AI - I'm writing creating genuinely useful programs I could never have dreamed of.

Let's not be totally down on AI!

I have an idea of an app I'd like to make, but when I have the time and headspace for it, I'm not going to vibe-code it. I want to understand the code I'm putting out. I've learned coding before, I can learn it again.

Like some have already commented, I have mixed feelings about this.

In one hand, discovering genuinely good apps will be harder, because you won’t be able to know whether an app is code-vibed, containing many bugs that the “code prompter” doesn’t know how to properly solve, or has been written by a coder even if it’s with the help of AI. No way to know unless you already know the dev.

But on the other hand… like user @CalMin said, this tools, that are getting better and better at coding, allow us, mere mortals that never were able to learn code, to easily make small programs or scripts.

Let me share an example. There’s been a software niche that wasn’t covered until a few months: an Obsidian-like knowledge management software, with local LLM integrated RAG system. Most of the local RAG systems I know are based on complex configuration that involve several software components.

So there’s a guy (or gal, idk) that six months ago started building this app for macOS, Canto, using AI to develop it much faster, quickly implementing new features in few days. I haven’t tried it yet, so I still don’t know how polished it is, but if it wasn’t for AI-assisted development, apps such as this one wouldn’t exist.

I accept that AI is a powerful tool, and in the right hands can be fantastic. Like you said though, how can we tell which apps are good or not? How can we see the difference between a vibe-coded slop-fest that is a pretty collection of vulnerabilities, an app that has been vibe-coded with some forethought that only has a few vulnerabilities, and will get hard to maintain over future versions, or a very thoughtfully made app that has had AI used in its creation, but is quality all the way through?

 

[Mac Fly (film)]

Here comes my first fart app.

Guideline 4.3(b):

"Don’t submit apps that are indistinguishable from what’s already widely available. Opportunistically creating variants of existing app categories or popular apps degrades App Store discovery, reduces overall app quality, and harms both users and developers. [...] Other kinds of apps, such as drinking games, Kama Sutra, fart, and burp apps, are mediocre, low-quality, or low-effort and do not add value to the App Store. Repeated submissions of this kind may lead to removal from the Apple Developer Program."

 

[svish]

Will be interesting to see as it develops and evolves over time. Can definitely help in building apps.

 

[syklee26]

I will create a game app where the goal is to fire Eddie Cue

 

[ry-guy]

Here comes my first fart app.

I know you’re joking but Apples real guidelines were updated to say they basically won’t allow fart apps anymore haha

 

[MilaM]

Brace yourselves for an avalanche of vibe-coded slop.

It's already happening. The cadence of new app releases (mostly copies of already existing apps and pet projects) has increased significantly over the last year. Many developers release them for free or just ask for modest donations outside the app store.

It will be interesting to see what happens next. Are we going to see additional demand, or will existing publishers have to live with less revenue?

 

[everlast3434]

Hi everyone. Lead developer here! I have more than 8 minutes of experience vibe coding. At Fart Universe, we’re building the worlds best fart apps. As the lead developer, I can tell you we’re creating amazing things. Thanks Clud!

 

[djplong]

Full Disclosure - I work for the Air Force.

I'm of two minds of it. Like 'fire' and 'knives', there are good and bad sides.

This can enable people to write at least simple apps for their own uses (Fire: Cooking, keeping warn in the cold). I think most people can agree there's no harm in this. Now, the 'barrier' would be setting up Xcode. Me? I've been a software architect/engineer since my days as a junior programmer back when *Ford* was President.

Then there's the other side. The side that replaces humans with Ai coding and bugs get through (Fire: Burns people)

There was a study done here, recently, about the code generated by AI. We found that, for every 800 lines of code, 700 was useless slop. Not useful in the least. Even a human appendix has a use case (storing bacteria). Only 100 lines of code did the actual.

I'm working with someone in another area of our division and their conclusion is that Claude and similar projects are sufficient for small programming tasks with human oversight. The opinion included a statement that the most reliable use would be for small tasks, such as those given to junior programmers.

Think about that. We may be starting to cut off the 'breeding ground' of good coders. If you can't get a job as a junior programmer, you have no "on ramp" to gaining the experience that you need to take on bigger projects and, eventually, lead teams. It's like baseball team owners deciding that A-level baseball isn't useful enough and eliminating that level of play. That will, before too long, affect the major leagues.

I'm nearing retirement. Looking back, based on the experiences I have - this could be disastrous. Corporations will not be content with non-techies coding their own apps. They're going to slash staff because of having to "make next quarter's numbers". This has had AWFUL effects in so many industries over the past 40 years. One of the clearest examples was the supply chain catastrophe after COVID - and we're still feeling the after-shocks from that.

 

[dannys1]

Until you read the things Anthropic's leaders say about their fellow human beings.

I don't like humans either.

 

[ducknalddon]

You laugh but I published an app to the app store last month and I’m definitely not a developer.

I've been writing code for nearly fifty years and I'm wondering if I'll bother anymore.

 

[ducknalddon]

Whilst there maybe lots of slop out there, professional developers shouldn’t underestimate the power and advantage AI can give them as it has done me.

It gives you no advantage because everybody else has access to the same tools.

 

[leifp]

It gives you no advantage because everybody else has access to the same tools.

That’s a bit like saying having a hammer is not an advantage… if I want to hammer in some nails, it’s a solid advantage. Beats fists or rocks…

 

[manhattanboy]

No before Claude had to take over the iOS simulator app, requiring accessibility permissions etc. Claude would take over your computer to do this. Now all of this runs inside the Claude desktop app.

The takeover route can still accomplish way more, and is not limited to iOS.

I don’t think many know how proficient the top models have become. Anything a human can do on a computer, by the end of this year AI will be able to do it better, if it doesn’t already.

 

[Populus]

How can we see the difference between a vibe-coded slop-fest that is a pretty collection of vulnerabilities, an app that has been vibe-coded with some forethought that only has a few vulnerabilities, and will get hard to maintain over future versions, or a very thoughtfully made app that has had AI used in its creation, but is quality all the way through?

Yes. Totally legitimate question that I can’t answer unless you know the developer and his previous work. It’s a legitimate worry that I also have.

However, just to put a positive note on all of this, AI could act as a code reviewer to check whether the software have any weak points, potentially causing bugs or known security vulnerabilities. Or even the end user can check it by themselves if the app is open sourced.

 

[User 6502]

Brace yourselves for an avalanche of vibe-coded slop.

The App Store has been full of human-coded slop from the start. Or do you think fart apps and flashlight apps were programming masterpieces? Slop is slop no matter who authored it. Very few apps are actually good and it will be the same for the one produced by coding agents. Most will suck and a few will be good.

 

[MilaM]

The App Store has been full of human-coded slop from the start. Or do you think fart apps and flashlight apps were programming masterpieces? Slop is slop no matter who authored it. Very few apps are actually good and it will be the same for the one produced by coding agents. Most will suck and a few will be good.

True. But it's also becoming increasingly difficult to assess which ones are good and which ones are the bad apples. They all look very polished with pretty screenshots, homepages and extensive AI generated docs.

 

[mielie]

Full Disclosure - I work for the Air Force.

I'm of two minds of it. Like 'fire' and 'knives', there are good and bad sides.

This can enable people to write at least simple apps for their own uses (Fire: Cooking, keeping warn in the cold). I think most people can agree there's no harm in this. Now, the 'barrier' would be setting up Xcode. Me? I've been a software architect/engineer since my days as a junior programmer back when *Ford* was President.

Then there's the other side. The side that replaces humans with Ai coding and bugs get through (Fire: Burns people)

There was a study done here, recently, about the code generated by AI. We found that, for every 800 lines of code, 700 was useless slop. Not useful in the least. Even a human appendix has a use case (storing bacteria). Only 100 lines of code did the actual.

I'm working with someone in another area of our division and their conclusion is that Claude and similar projects are sufficient for small programming tasks with human oversight. The opinion included a statement that the most reliable use would be for small tasks, such as those given to junior programmers.

Think about that. We may be starting to cut off the 'breeding ground' of good coders. If you can't get a job as a junior programmer, you have no "on ramp" to gaining the experience that you need to take on bigger projects and, eventually, lead teams. It's like baseball team owners deciding that A-level baseball isn't useful enough and eliminating that level of play. That will, before too long, affect the major leagues.

I'm nearing retirement. Looking back, based on the experiences I have - this could be disastrous. Corporations will not be content with non-techies coding their own apps. They're going to slash staff because of having to "make next quarter's numbers". This has had AWFUL effects in so many industries over the past 40 years. One of the clearest examples was the supply chain catastrophe after COVID - and we're still feeling the after-shocks from that.

AI could barely string a function together 2 years ago. A year ago half the code I got from ChatGPT wouldn't even compile. Now tools like Claude Code and Codex iteratively work through the problem, compiling and testing as they go. You may be correct right now but the trajectory clearly shows AI coding is getting exponentially better.

People who continue to bang on about how experiences software engineers are better are, IMHO, just sticking their head in the sand. Soon there will be zero need for engineers to write code.

It won't be disastrous either. Yes Corporations could 'slash staff' to hit a quarters target, but at the same time someone in their bedroom can put together a tool/app that could undermine that entire corporations product line. AI is actually taking power away from big business and handing it to anyone with a great idea.

Its going to be a really exciting future I think!

 

[User 6502]

True. But it's also becoming increasingly difficult to assess which ones are good and which ones are the bad apples. They all look very polished with pretty screenshots, homepages and extensive AI generated docs.

If they are very polished and extensive docs who cares if they were authored by a human or a machine? If it’s good it’s good.

 

[MilaM]

If they are very polished and extensive docs who cares if they were authored by a human or a machine? If it’s good it’s good.

A polished interface used to be a good proxy for how much work a care developer put into designing it. Now the AI cranks out a beautiful UI, but you never know if the application logic is also well-designed. Doesn't matter so much when it is a "fart app". Could be problematic when the product is a password manager.

 

[Frantisekj]

I guess Im a developer now.

watch Apple posting 300+ App reviewers job postings 😂

 

[turbineseaplane]

It sure seems to me like the price charged for software should be going way down with much less human involvement in the creation.

 

[Enlightened Doggo]

I wonder if local models can do it too. I only run local models so would be pretty cool.

I've run local models through Claude Code with mixed results. It works but the harness is clearly optimized for running "swarms" against a cloud, where locally you generally just want to be running one model at a time.

 

[Agent007]

I've run local models through Claude Code with mixed results. It works but the harness is clearly optimized for running "swarms" against a cloud, where locally you generally just want to be running one model at a time.

What's a harness? What's a swarm? A swarm of what? Against what cloud? Honestly!