[xp84]

More and more, I wonder if projects like this (I assume they're made possible 'all of a sudden' because of AI coding tools) are going to eventually breathe new life into dozens of "obsolete" platforms long ago deemed too costly to keep supporting. Maybe we'll be able to daily drive OS9 with modern browsers and bugfixes. Or for those of us less masochistic, say, Windows 2000 or NextStep!

[itomato]

One of the first things I did with ChatGPT was feed it source for the existing OHCI USB Host Adapter driver to make it work with more common UHCI hardware.

There are many people working to extend NeXT, Openstep and even Rhapsody. Not to mention emulation projects and deep dives into XNU source, early Darwin source and shoehorning kernels and libraries into various places.

The Epson NX was purpose-built to run NeXTStep, for example, and bringing support for machines like that and the Canon object.station into 86box is a pet project of mine.

Join us https://www.nextcomputers.org/forums/index.php?topic=5286.ms...

[User23]

That reminds me of how I had had to cross compile a Linux kernel with a special USB patch to install a system on my Bondi Blue iMac.

[CursedSilicon]

At least in retro spaces there's a lot of rejection of using AI to do things.

Half-Life (and the later Source engine games such as Half-Life 2 and Portal) were ported without LLM assistance however

(Source: I am in the Discord where the developer has been working and diligently posting updates on the project)

[Jackknife9]

Yeah, I’ve definitely seen this sentiment in the retro tech community.

I think hobbiests do not have that much free time to create these tools for retro computers. AI is a great tool for this, as it saves time so people can make projects that would otherwise not exist.

I remember arguing with someone on Reddit who said they’d rather not have the tool if it was made by AI, which I thought was a crazy take.

[userbinator]

At least in retro spaces there's a lot of rejection of using AI to do things.

I suspect the community will fracture, and then once the pro-AI side shows off more and more of the amazing things they accomplished, they'll grow while the anti-AI side will eventually dry up into a few luddites.

I'm old enough to have been on the anti-HLL side when I was handwriting Asm for the PC, but of course HLLs took over because they're "good enough" and became a massive force multiplier.

[turpentine]

Yet the dev has posted here on HN saying they used LLMs to do the port, but have a preference not to communicate their AI use.

[CamperBob2]

(Removed inaccurate remark, or rather, removed my reply to an inaccurate remark)

[srp-shay]

Hi there, I'm the developer of this port.

I did use AI to assist me in development and I'm not shy about admitting that. This was still a hands-on effort that took well over 40 hours of development time spread over a couple months. Much like other development tools, AI certainly sped up my development process and allowed to ship something in a couple of months using my free time.

AI is a tool and it isn't going anywhere, I've posted my thoughts about AI extensively in the PowerPC community. I don't proactively volunteer or plaster my readmes with notes about AI because it doesn't really matter.

Speaking for me personally I have no problem using tools in my toolbelt to solve problems faster in my limited free time.

[yowzadave]

I recommend you approach life in a spirit of curiosity—it can be interesting to learn about the different interests and motivations of other people!

[hoten]

"unsustainable" is an odd word to use for a hobby. They aren't trying to hit productivity goals. Maybe they just still like programming.

[CamperBob2]

I do, too, but I'm feeling increasingly awkward doing it by hand.

NGL, it's not an altogether-comfortable feeling. At least not yet.

[CursedSilicon]

What a bizarre take

Many people refuse to use LLM's for various reasons that I don't need to rehash here

[CamperBob2]

"Bizarre" is rejecting powerful tools on ... what, religious grounds?

It would be one thing if we were talking woodworking or pottery or another skilled craft, where your personal growth of talent with experience is evident in the product, resulting in something you can look back on fondly and point to with pride.

But moving some legacy code (which you didn't write) from one platform to another? What do you get out of doing that by hand?

[CursedSilicon]

You seem either deliberately or functionally unable to grasp the premise of why anyone would do this (or reject LLM's)

I suggest instead you go read up on why people dislike them and gain a more well-rounded and nuanced understanding of the topic

[CamperBob2]

And you seem to be operating on a faulty premise (as was I.)

[ColdStream]

One impressive one to me is the current effort to port Doom to the Neo Geo. Because of its unique hardware setup in that it is entirely sprite based and doesn't have a frame buffer, it is a very difficult challenge to take on. And yet most of the code on these projects is being done via AI and it works pretty dang well.

Considering these coding system probably have almost no idea of the Neo Geo's specifics and with very little reference code directly linked to the system to bounce off, that is dang impressive.

[russelg]

This isn't an all-of-a-sudden, the engine has been worked on for over 7 years now https://github.com/fwgs/xash3d-fwgs

[Coeur]

Or even create new platforms where we're not bound by platform owners' imposed artificial limitations.

[ColdStream]

So the entire free software movement? ;)

[cellularmitosis]

related: https://powerfox.jazzzny.me/

[user3939382]

The big blocker is actually TLS. Without it lots of tools including browsers are blocked. Now with AI I bet it is manageable, before that the effort required resulted in what you see, almost nothing.

[userbinator]

TLS is not really that much of a blocker, nor was AI necessary to clear that hurdle (albeit crypto is fundamentally "just maths" and thus very, very portable): https://news.ycombinator.com/item?id=36486512

[someperson]

I feel like the best Beatles songs are yet to be written.

Or alternatively best classic Simpsons episodes are yet to be made.

[dosisking]

> Or alternatively best classic Simpsons episodes are yet to be made.

Yes. This will happen.... when pigs fly

(It's still good, it's still good)

[gcanyon]

I think you're underselling where we'll get to in the next 5-10 years. I'm thinking:

Best Beatles song recorded between Rubber Soul and Revolver, composed mainly by McCartney but Lennon wrote the bridge and Ringo was feeling adventurous and Harrison was angry that day, and it was deemed not commercial enough to include on Revolver.

Best season 3 episode of The Simpsons, really showing the transition to Homer as the lead, written by Robert Cohen with significant musical input from Danny Elfman and featuring seven characters loosely representing the deadly sins, with Homer as Gluttony of course.

[soulofmischief]

I think people seem to be downvoting this because it's not something they like whether true or not, however I think it's a thought-provoking perspective which warrants some consideration.

[nonethewiser]

I think there is real insight but it kinda relies on a rhetorical trick. AI could arguably create a simpsons episode. It absolutely could not make a Beatles song. The Beatles can only make a Beatles song, and they do not exist.

[xamuel]

The Beatles absolutely still exist! They published a new single in 2023 ("Now and Then") and they surgically injected slices from archive footage into new footage for the music video to make it look like young Lennon and Harrison perform alongside elder McCartney and Ringo.

[anthk]

There's already Netsurf with beter JS support than the Duktape one under Unix, Windows and even RiscOS.