This has been an unexpected outcome of agentic coding for me.
I recently ‘rescued’ [0] one of my favourite video game series, No One Live Forever, which is kind of lost to IP hell. It’s been an enjoyable experience, I learnt a lot, and it’s feels great to have revitalised the game.
I'm the opposite... if it's too easy, it's just not fun or special to me anymore.
And the amount of people using AI to flood the internet with new bugs and exploits means it's inevitable that maintainers will also begrudgingly have to resort to using AI just to keep up.
I'm with you, I wouldn't have expected this a year ago. While it's been awesome for the hacker ethos, I worry that it might be too much too quickly. Commercial software will push to online-only business models to protect their IP from decomplication. For legacy/retro software, this couldn't be better.
One angle I don't see anyone talking about (perhaps because I'm mistaken) is that this is terrible news for Nvidia. AMD also makes great hardware, but their software is often seen as holding them back. A successful open-source CUDA port/decomp could devastate the AI pecking order.
I don't think commercial software will survive something like this. There’s no reason to subscribe to any software when there are open-source alternatives for everything. The only way to make money with software in the future is by selling additional services that require computing power. For example, if Adobe and Microsoft were to make their software open source, they could still continue to make money through AI subscriptions. The software itself will no longer have any monetary value.
Server components won't do. The model for protecting their IP will be streaming games. Maybe AI would be able to produce the game just from a video, but unless the video includes every single aspect of the game, that would be a lossy process.
It’s not technically or financially viable. AAA games are already too expensive to create, and this increases the complexity by an order of magnitude. And that’s before the latency problems that make the game less fun, and therefore more likely to bomb.
I bet they would be able to figure out a lot through game guides over the years, but not sure how popular those are. I play a lot of search action games and still look things up a lot.
I wouldn't waste so many resources and time by every remotely interesting thing (down to whatever someone may remember their older brother playing when they were young) being "made" from scratch by 50 different people, who usually don't communicate with and often don't even reference each other. "Oh hi guys, so I happened to have made this thing..." is kinda getting old.
> Commercial software will push to online-only business models to protect their IP from decomplication.
uhm.. My dude, AI can make entire apps from just a couple English sentences, it'll handle cloning an entire online-only app it can see (and sniff)
And frankly, who even feels sorry for Adobe, Microsoft, Apple, Facebook, Google etc?
They made trillions of dollars and looked down on their users for decades. Google is still anal about dark mode if you're not signed in. YouTube still hinders iOS's picture-in-picture feature. Facebook still doesn't have iPad-native versions of their apps. There's so much shit in apps with billions of users that hasn't been fixed for years.
And they will keep making money hand over fist, while anyone who worked however one long on something can have it snatched up within 5 minutes of them releasing it to the world. And then 10 more people clone that. We could enter some kind of dark age where people are more likely to try to keep even their ideas secret. If nobody gets any recognition, and putting in the work for others to clone it is just being a sucker, then people will stop doing that. And companies will simply yweigh what they invest into vs. what they can defend/attack with lawyers, or patent, or whatever. And then Google, Facebook, Apple, Microsoft and others can ooze over the space where there was society and community, even more so than they already did in the last quarter century.
> Google is still anal about dark mode if you're not signed in. YouTube still hinders iOS's picture-in-picture feature. Facebook still doesn't have iPad-native versions of their apps. There's so much shit in apps with billions of users that hasn't been fixed for years.
They're not selling apps, alternative apps do nothing to them. Sure they hate it, just like Twitter hates xcancel, but they could give them amnesty and servers instead and still would never be meaningfully harmed by a mere viewing tool of their data.
AI can also fix OS compatibility issues for older games:
I had Aviary Attorney lying around (great game btw, even if you don't like Phoenix Wright) but it wouldn't run on newer macOS versions. I tried asking Astra on a whim, looked away, and within a few minutes it was on screen!
Turns out "The game used 32-bit Node-WebKit, an old combination of Node.js and Chromium that modern macOS cannot run. Its actual game content was HTML, JavaScript, images, and audio, so replacing the runtime was feasible."
I recently played a little bit with quite naive and simple prompts like "convert this DOS game to run using SDL without dosbox" and in quite a few cases it made a good plan for porting using a custom tailored x86 16bit real mode emulator and then hooking IO on modern foundation and then having it port to some modern language (a few experiments in C++, some Rust) For some games I didn't get to a working state, but for many, especially around transition to VGA, where EGA/CGA had been supported too, they seem to have somewhat isolated and simple graphics for this to work with an okayish amount of tokens.
Not the OP but I've tried something like this for a couple things. It costs a $100 monthly plan for about 2 weeks or so, so $50 about? (which is equivalent to about $4500 with API pricing, if you were using Claude, or maybe $100 if you are using something with less grossly inflated pricing).
There's a lot of activity, but not a lot of it is from people who know what they're doing. Photocraft is less usable now than it was on release, which is impressive. One of the latest releases (I believe 0.11) had a bug where choosing a color on the color palette would hard lock Windows and require a physical restart.
Can you tell me more? Just really curious about this approach, and am not a Photoshop expert or even user. How usable was Photocraft on release, and what were significant gaps if any - outright bugs, missing features, subpar implementation of features? How did it become worse with time?
This has been an unexpected outcome of agentic coding for me.
I recently ‘rescued’ [0] one of my favourite video game series, No One Live Forever, which is kind of lost to IP hell. It’s been an enjoyable experience, I learnt a lot, and it’s feels great to have revitalised the game.
0: https://github.com/nolffan
I'm the opposite... if it's too easy, it's just not fun or special to me anymore.
And the amount of people using AI to flood the internet with new bugs and exploits means it's inevitable that maintainers will also begrudgingly have to resort to using AI just to keep up.
I'm looking to leave the industry entirely soon.
I'm with you, I wouldn't have expected this a year ago. While it's been awesome for the hacker ethos, I worry that it might be too much too quickly. Commercial software will push to online-only business models to protect their IP from decomplication. For legacy/retro software, this couldn't be better.
One angle I don't see anyone talking about (perhaps because I'm mistaken) is that this is terrible news for Nvidia. AMD also makes great hardware, but their software is often seen as holding them back. A successful open-source CUDA port/decomp could devastate the AI pecking order.
I don't think commercial software will survive something like this. There’s no reason to subscribe to any software when there are open-source alternatives for everything. The only way to make money with software in the future is by selling additional services that require computing power. For example, if Adobe and Microsoft were to make their software open source, they could still continue to make money through AI subscriptions. The software itself will no longer have any monetary value.
Server components won't do. The model for protecting their IP will be streaming games. Maybe AI would be able to produce the game just from a video, but unless the video includes every single aspect of the game, that would be a lossy process.
It’s not technically or financially viable. AAA games are already too expensive to create, and this increases the complexity by an order of magnitude. And that’s before the latency problems that make the game less fun, and therefore more likely to bomb.
I bet they would be able to figure out a lot through game guides over the years, but not sure how popular those are. I play a lot of search action games and still look things up a lot.
I’d get decompiling whilst the tokens heavily subsidised. Adobe should be pretty scared.
I wouldn't waste so many resources and time by every remotely interesting thing (down to whatever someone may remember their older brother playing when they were young) being "made" from scratch by 50 different people, who usually don't communicate with and often don't even reference each other. "Oh hi guys, so I happened to have made this thing..." is kinda getting old.
> Commercial software will push to online-only business models to protect their IP from decomplication.
uhm.. My dude, AI can make entire apps from just a couple English sentences, it'll handle cloning an entire online-only app it can see (and sniff)
And frankly, who even feels sorry for Adobe, Microsoft, Apple, Facebook, Google etc?
They made trillions of dollars and looked down on their users for decades. Google is still anal about dark mode if you're not signed in. YouTube still hinders iOS's picture-in-picture feature. Facebook still doesn't have iPad-native versions of their apps. There's so much shit in apps with billions of users that hasn't been fixed for years.
Fuck them.
And they will keep making money hand over fist, while anyone who worked however one long on something can have it snatched up within 5 minutes of them releasing it to the world. And then 10 more people clone that. We could enter some kind of dark age where people are more likely to try to keep even their ideas secret. If nobody gets any recognition, and putting in the work for others to clone it is just being a sucker, then people will stop doing that. And companies will simply yweigh what they invest into vs. what they can defend/attack with lawyers, or patent, or whatever. And then Google, Facebook, Apple, Microsoft and others can ooze over the space where there was society and community, even more so than they already did in the last quarter century.
> Google is still anal about dark mode if you're not signed in. YouTube still hinders iOS's picture-in-picture feature. Facebook still doesn't have iPad-native versions of their apps. There's so much shit in apps with billions of users that hasn't been fixed for years.
They're not selling apps, alternative apps do nothing to them. Sure they hate it, just like Twitter hates xcancel, but they could give them amnesty and servers instead and still would never be meaningfully harmed by a mere viewing tool of their data.
I've always meant to play this, had actually asked GPT about doing a decomp jsut the other day, and it surfaced yours. Nicely done
Give it a go and file me some bugs. :D
That's awesome!
https://github.com/nolffan/nolfsrc
Thanks!
AI can also fix OS compatibility issues for older games:
I had Aviary Attorney lying around (great game btw, even if you don't like Phoenix Wright) but it wouldn't run on newer macOS versions. I tried asking Astra on a whim, looked away, and within a few minutes it was on screen!
Turns out "The game used 32-bit Node-WebKit, an old combination of Node.js and Chromium that modern macOS cannot run. Its actual game content was HTML, JavaScript, images, and audio, so replacing the runtime was feasible."
I recently played a little bit with quite naive and simple prompts like "convert this DOS game to run using SDL without dosbox" and in quite a few cases it made a good plan for porting using a custom tailored x86 16bit real mode emulator and then hooking IO on modern foundation and then having it port to some modern language (a few experiments in C++, some Rust) For some games I didn't get to a working state, but for many, especially around transition to VGA, where EGA/CGA had been supported too, they seem to have somewhat isolated and simple graphics for this to work with an okayish amount of tokens.
LLM demonstrated once again how open source was just the right thing both from the ethical, technical and market pov.
Companies relying in selling blobs have just no value now. The ones selling services and using open source models are just ahead of times.
If you're interested in this, we're systemically porting the whole 8-bit game catalog to JavaScript over at https://gamesexplained.com/
and the OSS kit to do it is https://github.com/gamesexplained/gamesexplained/tree/main/k...
I've converted three large programs written in macro assembler to C, so it could be ported to another platform. It was extremely tedious.
(Article is from 2003 but I'll comment on the sentiment)
A new era of ownership is here.
I ported a closed-source game I'm a huge fan of from mac to iOS so I could play it on the go. No plans to release it; just sweet to play how I want.
The limit is your imagination!
How much does it generally cost? Like ballpark
Not the OP but I've tried something like this for a couple things. It costs a $100 monthly plan for about 2 weeks or so, so $50 about? (which is equivalent to about $4500 with API pricing, if you were using Claude, or maybe $100 if you are using something with less grossly inflated pricing).
The game was $10 or so on GoG. The Claude sub is the same one I use for everything.
(2003)
All my monthly software subscriptions have been going bye-byte. No more Bookmap, Adobe, or Quickbooks!
Did you implement bank sync for you quickbooks replacement?
No, been using it for a small business with not enough volume to need it.
I’m curious what Adobe software were you using that’s replaced, and with what?
probably https://getartcraft.com/ It's really shaping up, from the activity I see in Github.
There's a lot of activity, but not a lot of it is from people who know what they're doing. Photocraft is less usable now than it was on release, which is impressive. One of the latest releases (I believe 0.11) had a bug where choosing a color on the color palette would hard lock Windows and require a physical restart.
Can you tell me more? Just really curious about this approach, and am not a Photoshop expert or even user. How usable was Photocraft on release, and what were significant gaps if any - outright bugs, missing features, subpar implementation of features? How did it become worse with time?
I thought that was the whole point of the NT kernel, to prevent bugs like that from taking down the entire machine?
How is it able to lock up Windows entirely?
Please no snarky anti-MS comments.
Did you read the article?
The authors should have a look at: https://news.ycombinator.com/item?id=50152466
You didn't explain why?
the authors of a 2003 article?