it started (AI)

If something like this is possible with AI and Rust, maybe this can finally solve FreeBSD's wifi bugs? And maybe iron out the GPU issues while we're at it? And maybe the ROCm?

This can also be an encouraging sign for RedoxOS?

What's crazy is frankly the compilation and testing - the whole CI/QA feeldback loop. A Jenkins instance would probably go berserk just trying to keep up.
 
here is claude fixing it
1791803520160.png
 
Are you all reading what it says on that link?

Reactions were sharply divided between awe at AI-assisted engineering velocity and skepticism from seasoned toolmakers:


  1. The Optimists: Seeing seven native Rust creative tools appear overnight without telemetry or monthly subscriptions felt like an open-source triumph against Adobe’s corporate rent-seeking.
  2. The Pragmatists: Power users immediately tested realistic workflows and encountered the classic Ninety-Ninety Rule of software engineering (the first 90% of the code accounts for the first 90% of development time; the remaining 10% takes the other 90%). Japanese designer @fujix noted that PhotoCraft completely failed to accept Japanese text input, and canvas panning stuttered under complex layout PSDs.
To the author’s credit, the repository openly admits that “vibe-coding” initial architecture creates shallow coverage: early alpha testers discovered 214 keyboard shortcut failures, offset text carets, and sticky crop handles that naive unit tests had missed. Wiring 626 menus with an LLM is fast; perfecting thirty years of human motor interaction requires patient, disciplined engineering.
 
Crazy future is coming.
Update 1: The ArtCraft project has also generated code in Rust and released initial versions of the WordCraft word processor, GridCraft spreadsheet system, and DeckCraft presentation editor, implemented in the style of Microsoft Word, Excel, and PowerPoint. Builds are available (1, 2, 3) for Linux, macOS, and Windows.
Update 2: CADCraft (an AutoCAD-style CAD system) and SoundCraft (an audio editor modeled after Avid Pro Tools) have been released.
The era of traditional programming is drawing to a close. That much is obvious. The question is: to what extent will we have to deal with the fallout, and on what scale? Bill Gates, for instance, says that AI-driven cyberattacks will soon sweep across the globe. It is possible that these AI clones of popular programs will also be linked, in some way, to that AI-fueled chaos.
 
I find it odd/annoying that a 100pct ai derived tool is in the ports, but when I ask here if a few tools I wrote personally are worth entering into ports I get a resounding "no" (but I suspect the ai didn't ask; they just did).
 
I find it odd/annoying that a 100pct ai derived tool is in the ports, but when I ask here if a few tools I wrote personally are worth entering into ports I get a resounding "no" (but I suspect the ai didn't ask; they just did).
Entry into the Ports Collection is relatively easy, but please be aware that unmaintained ports do get dropped.

As an example, a really nice subtitling program that I like (multimedia/aegisub) got dropped. According to FreshPorts, that's because it fails to build. And that port has been dead for awhile now.

If you read the Porter's Handbook carefully, it tells you to play with your own copy of the Ports tarball first, make sure your stuff works correctly within that system, and then submit an application to be admitted into the official Ports Collections. If you do that, you basically sign up to be a maintainer. And yeah, if you're the maintainer, you'll be the one facing the music of getting dropped if you neglect the port or don't take rules of the road seriously.

I actually take some pride in spending time in rabbit holes noticing that kind of stuff. If someone needs help of an AI to even notice that reality of life as a software maintainer can burn, let alone know how to handle that - go play in a jail(8)...
😩
 
  • Like
Reactions: drr
There is a strong tendency not to take the "free puppy" variant of free stuff in the FreeBSD project. It will get stronger in the AI age.

Any links to those rejected ports?
 
If you read the Porter's Handbook carefully, it tells you to play with your own copy of the Ports tarball first, make sure your stuff works correctly within that system, and then submit
That goes without saying (you shouldn't have to be told to do that which is common sense). And, I did that long before I asked here if what I did was worth creating a port for (I offered a choice) because I automated pkg setup with my "git server" (it builds and packages my repos)--I stopped building an automated port configuration thing for my stuff after the "answer was no". My point was that ai manuf tools are showing up ports without any consideration other than "reach" (at least I asked a population/userbase).

I'd be more interested in a conversation if you honestly believe that a developer should NOT test their own code before and/or a statement like that isn't an overstatement.

The implication that an ai has/could have a lifespan longer than a few moments is an interesting one though.
 
That goes without saying (you shouldn't have to be told to do that which is common sense). And, I did that long before I asked here if what I did was worth creating a port for (I offered a choice) because I automated pkg setup with my "git server" (it builds and packages my repos)--I stopped building an automated port configuration thing for my stuff after the "answer was no". My point was that ai manuf tools are showing up ports without any consideration other than "reach" (at least I asked a population/userbase).

I'd be more interested in a conversation if you honestly believe that a developer should NOT test their own code before and/or a statement like that isn't an overstatement.

The implication that an ai has/could have a lifespan longer than a few moments is an interesting one though.
It's not out of question for an alpha version to show up in Ports as a means of reaching more users - provided the maintainer actually minds the feedback and updates the port regularly, with bugs/crashes/errors fixed in each subsequent release/commit. Well, some well-known projects (like KDE 6) spend time in Area51 repos before getting moved to the official Ports Collection. And I have compiled Area51 code before - because I was willing to live with the consequences of messed-up dependency specs and the like.
 
an agent cannot "live" long enough to care what happens in the future. at best, a subsequent agent can be motivated to pick up where the last left off. what motivation is there for a non developer to maintain abi (with a language that does not even have a formal spec, no less)? what cost was there for an actual human to develop v a non developer spinning up an instance (an actual human with actual stake in a project has better chance of survival)?

compiling "bleeding edge" is no different than compiling a project in early phases so I'm not quite sure what your point is. I would venture a guess that it could be safe to assume that +90% of the given userbase of a *nix system has compiled alpha/bleeding edge at some point.
 
Any links to those rejected ports?
No. I didn't make it as far as submitting a port. I made a program to do X, asked users here if they thought it was worth making a port for. Most I suspect never even read my project's readmes (I would typically get a response like "ansible can do that" or some other "out of left field" reply) but I did get at least one flat-out "no" response (I asked for a reason but never received one) for one project of mine. ...my point is that I asked "a userbase" (whether or not this is the proper place to ask or not is not really relevant) where as these ai driven tools are only seeking the furthest reach as possible (for what, I do not know--my motivation was obviously to help others with similar needs).
 
an agent cannot "live" long enough to care what happens in the future. at best, a subsequent agent can be motivated to pick up where the last left off. what motivation is there for a non developer to maintain abi (with a language that does not even have a formal spec, no less)? what cost was there for an actual human to develop v a non developer spinning up an instance (an actual human with actual stake in a project has better chance of survival)?

compiling "bleeding edge" is no different than compiling a project in early phases so I'm not quite sure what your point is. I would venture a guess that it could be safe to assume that +90% of the given userbase of a *nix system has compiled alpha/bleeding edge at some point.
my point was, the very admission process into Ports Collection is most likely NOT outsourced to an AI. Generating code, that was. But spreading the word and seeking out a userbase to get feedback on very initial results - that was most likely the case that was made when applying for admission into the Ports Collection. And even if the port were to be rejected by the Portmaster, not a big problem for the author - the stuff is on Github, runs on Linux and Windows, author can get feedback that way, too. Even if code is AI-generated, there's still user feedback to contend with. If you don't pay attention to user feedback and don't release fixes based on that feedback (as in, tell the AI to code up the fix based on user feedback), you'll be forgotten and ignored into oblivion.

And I'd venture that you're way off on your assumption that >90% of a given userbase of a *nix system have compiled alpha/bleeding edge. Just consider how widespread Linux is, how widespread Android is. My guess would be that less than 15% of the users even know what it means to compile software at all, let alone compile alpha/bleeding edge stuff. Most users stick to available packages and do distro-hopping if they don't like the packages of one distro. I was like that, too, but I walked the thorny path to knowledge.
 
AI Agents live long now. The OpenAI vs. Huggiingface incident clearly showed that the second thing the agents did is gang up, specifically also to have their work continue after one particular's agent resources run out.
 
I installed the PhotoCraft clone of Photoshop on my Win10 workstation.

Adobe refused to allow my lifetime license for CS6EE to ever activate again, and told me to buy a subscription.

PC is vibe coded in RUST and seems to work ok. It feels a bit slow compared to CS6EE but I have just scratched the surface this morning.

The predatory subscription environment looks to have its days numbered.
 
I installed the PhotoCraft clone of Photoshop on my Win10 workstation.

Adobe refused to allow my lifetime license for CS6EE to ever activate again, and told me to buy a subscription.

PC is vibe coded in RUST and seems to work ok. It feels a bit slow compared to CS6EE but I have just scratched the surface this morning.

The predatory subscription environment looks to have its days numbered.
I never bought a photoshop license, always looked for alternatives, GIMP was actually not bad. And within US, the predatory subscription practices got attention of federal government agencies in DC a couple years ago. And, if you wanna do something just for fun from time to time, there are online services for that. Some try to charge a fee, most are free with limitations, like the Drake meme generator, or "manul generator" (google those).
 
back to something more resembling on topic...
for me, this ai stuff has turned programming into something more akin to fishing. I now: go out, catch a fish, eat fish. vs go out, catch fish, share fish. This simplified the process quite a bit because I don't have to worry about licensing or timely bug fixes or anything else. my (previously planned) public projects are now, privately, several versions beyond the public ones and I have several new private only projects that should never be seen by anyone.

My code can be automatically built and automatically distributed all from within my infrastructure and I don't have to worry about anyone else's infrastructure/needs. github (aka: public code hosting) is dead for me.

yay for ai.
 
I have to maintain a Win10 workstation for those apps I bought (digital audio workstation) and Excel with VBA macros.
I'm building a Xeon based hotrod new workstation, but found my licensed Photoshop will not activate.

Yes, I did grab GIMP but it is radically different from CS6EE.
The PhotoCraft clone is remarkably similar to my CS6.

I do a lot of female retouching work, so those few tools I use in CS6 are critical.
I may be forced to keep my existing i7-870 workstation alive and separate simply to keep CS6EE.
 
I have to maintain a Win10 workstation for those apps I bought (digital audio workstation) and Excel with VBA macros.
I'm building a Xeon based hotrod new workstation, but found my licensed Photoshop will not activate.

Yes, I did grab GIMP but it is radically different from CS6EE.
The PhotoCraft clone is remarkably similar to my CS6.

I do a lot of female retouching work, so those few tools I use in CS6 are critical.
I may be forced to keep my existing i7-870 workstation alive and separate simply to keep CS6EE.
Why wouldn't it activate?
 
Back
Top