I Vibecoded Morrowind, Duke Nukem, and Serious Sam Onto an iPhone

[ArjunKL]

I can't code, but with AI agents I built native iPhone ports of Duke Nukem, Serious Sam, and Morrowind, learning where vibecoding works, where it fails, and why human direction still matters.

Read the full article here.

I Vibecoded Morrowind, Duke Nukem, and Serious Sam Onto an iPhone

 

[Vixiy]

I don't recall making a mixtape or burning a CD being such an involved progress, nor did it ever cost me hundreds of dollars and an entire neighborhood's health and peace.

 

[Deezy]

Great read, thank you!

Funny how we have gotten to the point now where we say "I finished by project in less than a week (part time), but OH MY, I burned through a whole $100 worth of AI credits!" Long term, to have any reasonable project completed, by a non-coder, for $100 and some time is an incredible bargain.

The times, they are a-changin'!

 

[umeng2002]

This could have been done for free without AI.

 

[Dr Roboto]

I personally thought this was a great article. I applaud all the hard work the author put into to try this. I especially liked how the author "managed" the agents to get the job done. I am amazed at what was accomplished. Great job.

AI coding is here to stay, but as the author pointed out, it takes experience and "management" to do it right. This was not just throw it over the fence and let an AI system do all the work. The author was a deeply integrate part of the process even though they couldn't write code.

 

[TechSam]

Really awesome to see. Non-technical people wouldn't have a chance in doing this, some people don't have an engineer mindset or interest in the technical part. But some people are great designers, they know what they want and how to ask for it.

Myself, I have moved away from using Unreal Engine all together and gone into reverse engineering games. I like this idea of being able to port old Arcade Games from the 90s to a modern C++ codebase. I'm having a great time doing it.

 

[Lew Zealand]

This could have been done for free without AI.

Yet it wasn’t.
But it was done with AI.
Hmmm…

 

[Squid Surprise]

This could have been done for free without AI.

Not by 99.9% of people though... AI brings this to the masses... and it will only improve with time.

 

[FF222]

"I Vibecoded" - that's an oxymoron. If it's vibecoded, then it wasn't you doing the coding.

Also, all AI did was copy an existing code, and tweaked it with already existing modifications, if anything. It didn't create anything new.

And finally the copyright/licence on the code is completely invalid, as AI generated code can not be copyrighted, even if it's legal to begin with. Which is questionable.

 

[FF222]

Yet it wasn’t.

But it was done with AI.

No. It was done _by AI_. The "author" had really played no measurable role in this.

 

[BadThad]

Incredible use of AI! BRAVO!

I found the details fascinating as an AI user myself (for work). AI is a tool, nothing more, it's like a prodigy child and needs to be treated as such. I've seen it flip flop 180° only to later completely pivot back. It has to guided along to the path of success by a skilled and reasoning human.

People fear losing their jobs to AI - and that may be true for some positions. However, the demand for skilled AI users will create new positions. Humans will always have to direct it and validate the output. Professionals will have to learn to embrace it and use it or perish.

 

[godrilla]

Speaking of AI. Did you know that RenoxDX used AI and uses vibe coding from his own original 100 mods to create new mods to create DLSS 5 RenoDX mod? Here is a video of the interview...

He also does the proofreading, testing and verification.

 

[Squid Surprise]

No. It was done _by AI_. The "author" had really played no measurable role in this.

Really? Would it have been done had the author not USED the AI?

Do we take Napoleon’s military victories away because it was actually his soldiers who did the actual “winning”?

 

[Lew Zealand]

No. It was done _by AI_. The "author" had really played no measurable role in this.

LOL, no.

AI was used by the author as a tool to accomplish a goal which nobody else accomplished. AI did a lot of the gruntwork which is precisely what its is good at and why people use AI, the stuff that takes people too much time or is convoluted enough not to even try at all. After all someone already failed to make the Duke Nukem port without AI and here someone made it with AI's assistance.

AI is a tool but a person needs to use the tool properly to solve the problem.

 

[FF222]

LOL, no.

AI was used by the author as a tool to accomplish a goal which nobody else accomplished.

LOL, no. There were already several ports of Duke Nukem and Serious Sam that ran on iPhone. That alone makes the argument that this would have been some kind of goal accompished that nobody else has accomplished yet, completely false.

But that's not even what my argument was about - it's just your straw man or red herring. My argument was about who did the actual coding. And since the author merely prompted the AI to do something for him, and it was the AI that did the actual coding, it's not the author that coded the app. QED. Simple as that.

AI did a lot of the gruntwork which is precisely what its is good at

LOL, no.

AIs (LLms) are only good at doing things _that have been already done several times_. It's because they can only do things that they have been taught to, by previous instances of the same thing. They can't achieve something completely new, because that's not how they work.

AI is a tool but a person needs to use the tool properly to solve the problem.

Which is for not even true, but more importantly, not what the argument is about. And it doesn't even make a compelling straw man.

 

[FF222]

Really? Would it have been done had the author not USED the AI?

Well, _the author_ likely couldn't have done it without using AI. But that doesn't mean that nobody could have or nobody did it already. Hint: they already did.

But the more important distinction is that the author didn't do (the coding) it this way either. The AI did it. He merely asked the AI to do it for him. He did not do the coding.

You understand the difference between these words and concepts, don't you?

Do we take Napoleon’s military victories away because it was actually his soldiers who did the actual “winning”?

That depends solely on what part and role they played. Also, "victory" is not an activity. "Coding" is. I made a claim about the latter. Your question was about the former.

You understand the difference between these words and concepts, don't you?

 

[Squid Surprise]

Well, _the author_ likely couldn't have done it without using AI. But that doesn't mean that nobody could have or nobody did it already. Hint: they already did.

You claimed the author played no measurable role in this - my argument is they did.

But the more important distinction is that the author didn't do (the coding) it this way either. The AI did it. He merely asked the AI to do it for him. He did not do the coding.

A tool serves its master - and the master gets credit for its accomplishments. When you “turn on the lights”, do you credit the electricity, wiring, bulbs? Or the person who flipped the switch in the room?

Answer: you say “I turned on the lights”, not “electricity/wiring/Edison turned on the lights”.

You understand the difference between these words and concepts, don't you?

Apparently you don’t.

That depends solely on what part and role they played. Also, "victory" is not an activity. "Coding" is. I made a claim about the latter. Your question was about the former.

How was winning the battle of Austerlitz (as an example of one of Napoleon’s many military victories) not an “activity”?

His soldiers did all the physical labour for it obviously… but… pretty much every military historian gives the credit for the victory to Napoleon’s generalship - not the soldiers.

You understand the difference between these words and concepts, don't you?

I don’t think you do…

 

[tonylukac]

Video games were invented in the 1950s at high end universities, but we were some of those who made them accessible to lower end users at the University of Illinois at Chicago in 1976 on an HP 2000 mini computer and an IBM 370 mainframe. All were semi command line games. There was an enlarged Star Trek game, I had an AI sort of Star Trek game I took from high school, an email, and a Blackjack game on the HP. The mainframe had various games including Vistrek which had graphic like cursor control, and an AI Adventure game. I think my Star Trek game, called Trek73, went on to create Star Wars and Star Trek movies.

 

[guillejp]

Impressive

 

[trgz]

I've been coding on and off since '79. I recently used AI to knock me up a script and I was reasonably impressed.
When I left my development job in 2019 the cost of development was being offshored to cut cost, and whatever way you look at it this is simply a progression of that. This might have cost $100 but how much in labour time would it have been, and what cost that? A lot more than $100 by a human.
I'm kind of conflicted about the use of AI as I remember the frustration of people breathing down my neck, whilst happily coding but wanting, at least for my own satisfaction, of finishing a product.

 

[FF222]

You claimed the author played no measurable role in this - my argument is they did.

That's not an argument. That's a (false) claim.

A tool serves its master - and the master gets credit for its accomplishments.

A tool does not "serve" its master. A tool is used by its master. If the tool's "master" does virtually nothing, then whatever is achieved is not his accomplishment, but either the tool's, or of those who created the tool. You do not get to claim you've created a car, just because you drive it. You only create a car when you actually design and build it.

Your point is also moot, because courts have already established (which I also wrote in my very first comment) that users of AI tools can't copyright the thing the AI has created. That's legalese for "they do not get credit for it". Why? Because it was not them who actually created the thing. They merely ordered that to be created.

When you order a pizza in a restaurant, you do not get to claim to have created the pizza. You just ordered it. When you instruct an AI to create a program, you do not get to cleam that you've create the program. You merely told the AI to do that for you.

When you “turn on the lights”, do you credit the electricity, wiring, bulbs? Or the person who flipped the switch in the room?

Why do you keep diverting from the original point about who did NOT do the coding, over and over? Because even you know you have no actual argument? Thought so.

 

[Squid Surprise]

That's not an argument. That's a (false) claim.

Your claim? I agree - which is why I’m arguing with you

A tool does not "serve" its master. A tool is used by its master. If the tool's "master" does virtually nothing, then whatever is achieved is not his accomplishment, but either the tool's, or of those who created the tool. You do not get to claim you've created a car, just because you drive it. You only create a car when you actually design and build it.

Not how it works… the “master” gets credit regardless of how much “actual work” they did. The CEO of Ferrari gets to claim he is responsible for the new Ferrari - even though all the actual work was by engineers, laborers, etc..

Your point is also moot, because courts have already established (which I also wrote in my very first comment) that users of AI tools can't copyright the thing the AI has created. That's legalese for "they do not get credit for it". Why? Because it was not them who actually created the thing. They merely ordered that to be created.

Copyright law is not the be all and end all - and those laws are still in the process of being written and will almost certainly be altered in the years to come. Why, you might ask? Because lawmakers tend to always be years behind technological advances…

When you order a pizza in a restaurant, you do not get to claim to have created the pizza. You just ordered it. When you instruct an AI to create a program, you do not get to cleam that you've create the program. You merely told the AI to do that for you.

The head chef gets credit for it - even if a sous-chef did all the work

Why do you keep diverting from the original point about who did NOT do the coding, over and over? Because even you know you have no actual argument? Thought so.

I’m not deviating - re-read what I’ve posted (and you’ve replied to)…

 

[Arthurik_jan]

That's not an argument. That's a (false) claim.


A tool does not "serve" its master. A tool is used by its master. If the tool's "master" does virtually nothing, then whatever is achieved is not his accomplishment, but either the tool's, or of those who created the tool. You do not get to claim you've created a car, just because you drive it. You only create a car when you actually design and build it.

Your point is also moot, because courts have already established (which I also wrote in my very first comment) that users of AI tools can't copyright the thing the AI has created. That's legalese for "they do not get credit for it". Why? Because it was not them who actually created the thing. They merely ordered that to be created.

When you order a pizza in a restaurant, you do not get to claim to have created the pizza. You just ordered it. When you instruct an AI to create a program, you do not get to cleam that you've create the program. You merely told the AI to do that for you.


Why do you keep diverting from the original point about who did NOT do the coding, over and over? Because even you know you have no actual argument? Thought so.

Jesus. Stop being obtuse and picking fights, dude. Borderline childish behaviour.

AI is a tool. He used the tool, managed, corrected, basically guided it to make something like this possible. He went from no port of the three games in the beginning to having ported all three to iOS, full credit to him. It makes no difference if it had been done before or not.

Coding with AI is not a simple task, and the more complex the task, like the porting of a game, the more human input the agent requires. The author of the article has to prompt correctly, check results, test, roll back, completely restart from scratch at times, prompt/design/test UI, solve compatibility issues, troubleshoot, bug hunt, basically do tons of acrobatics to get a game even as old as Duke Nukem 3D working the way he intends to on a mobile device with a touch screen interface. How should he not be credited with the work? He did it, good on him. If I were a fan of any of these franchises I'd pay and be thankful to him for taking the time and making it happen.

 

[rhinofinger]

You should consider launching your eduke32 port on the App Store if the open source license permits. GenZD on the iOS AppStore ports GZDoom over, and the user provides their own files. I’ve played numerous Doom mods through it, including the fantastic Ashes 2063.

I believe Delta Touch on Android similarly run source port(s) of the Doom games.

Point being - providing eduke32 through the App Store may be feasible, as long as users provide their own files. If you need to provide files for Apple/Google testers, the shareware version of Duke Nukem might work for that. I don’t believe the other games have shareware versions, so I’m less sure if you’d be able to provide those

 

[AlmightyLoaf]

A tool does not "serve" its master. A tool is used by its master. If the tool's "master" does virtually nothing, then whatever is achieved is not his accomplishment, but either the tool's, or of those who created the tool. You do not get to claim you've created a car, just because you drive it. You only create a car when you actually design and build it.

It seems like you're bending over backward here to alienate the author from his work, as if he only uttered the godlike incantation "Let there be Duke Nukem 3D on iOS" and so it was. And if that was true then you'd have a point. But if you'd read the article, you'd understand that wasn't the case.

Your other interlocutors have made the point (which you haven't fully grappled with yet) that there are multiple roles which are all crucial to execute a successful project. Some are directly building the product, like game designers, artists, coders, and testers. Some are higher level, like product/project managers and executives. And some people fill multiple of these roles simultaneously depending on the team size.

But when the project is done, it's everyone's victory and everyone gets to claim they created the product. That's why the credits block doesn't only list programmers. In fact, top billing usually goes to creative directors that may have never seen a line of code. Because code by itself doesn't make a Good Game™... it's only valuable in service of something bigger.

"Good" is the product of asking the right questions, solving problems, iterating, understanding your users, and ensuring that the work is consistently high quality... all while not losing sight of the high-level vision along the way. AI cannot reliably do any of these things yet, which is why people are still necessary to keep the project moving forward until all of the pieces are in place.

By your logic, a ship captain can't take credit for a successful voyage because he only gave direction to the crew who carried out his orders. But I think we all understand that the captain's value is not him physically turning the wheel, but rather deciding where to go and how to get there, and how to deal with the problems that arise on the voyage.