Adobe is Screwed. So is every other software company

[Brian The Snail]

In the end, Linux is the way to go. Sadly, I'm stuck on windows because fucking Daz Studio won't port Daz to Linux (yet).

I've got a feeling that may be on somebody's list at this point. Over at Renderhub they have all sorts of vendors writing utilities and plugins for DAZ that probably should have been in DAZ 6 to start with. I think it's only a matter of time before somebody vibe codes the whole damned thing. That ArtCraft Suite? Has a complete Linux port. Adobe never bothered either.

 

[TailsWin]

You can use Claude Code to create your own GIMP-like app, with the features you want.
Only try that if you're technical, though, know about coding. (Unless you only want something fairly basic).

I make my own apps for where what I want doesn't exist, or I want to experiment and learn something how things work. I'm a micromanager so I have to get everything exactly how I like it.

Making a bitmap graphics editor from scratch sounds like a nightmare. Yea, you can get AI to write the code, but you still need to specify how you want things to work. E.g. if you don't like Gimp - are you able to describe what exactly you don't like and how to fix it? If you gonna just tell AI to "make Gimp but more user-friendly", you end up in the same situation. Because unless you specified everything in detail, you'll just get another new program with new UI you have to learn. So at that point, why bother? Just learn to use any of the myriad of existing apps.

Indeed if you just want something simple, then sure, vibecode your own, but as features get added, complexity shoots up and choices need to be made.

That said, maybe I'll fork Gimp at some point and add some stuff that I'm missing. Or maybe I'll try to decompile XnView since that's been my sore spot for 25 years that it's not open source lol. I wonder how all this will affect small devs that make closed source apps.

 

[TailsWin]

Now this is more interesting. People are de-malware-ing their spyware-riddled smart TVs, And reviving cloud devices that are no longer "supported" (so you should buy a new one).

 

[FantasyForge]

I've got a feeling that may be on somebody's list at this point. Over at Renderhub they have all sorts of vendors writing utilities and plugins for DAZ that probably should have been in DAZ 6 to start with. I think it's only a matter of time before somebody vibe codes the whole damned thing. That ArtCraft Suite? Has a complete Linux port. Adobe never bothered either.

A cool point, Mr Snail. I reckon a determined engineer using Claude Code could achieve porting rather quickly!

 

[FantasyForge]

I make my own apps for where what I want doesn't exist, or I want to experiment and learn something how things work. I'm a micromanager so I have to get everything exactly how I like it.

Making a bitmap graphics editor from scratch sounds like a nightmare. Yea, you can get AI to write the code, but you still need to specify how you want things to work. E.g. if you don't like Gimp - are you able to describe what exactly you don't like and how to fix it? If you gonna just tell AI to "make Gimp but more user-friendly", you end up in the same situation. Because unless you specified everything in detail, you'll just get another new program with new UI you have to learn. So at that point, why bother? Just learn to use any of the myriad of existing apps.

Indeed if you just want something simple, then sure, vibecode your own, but as features get added, complexity shoots up and choices need to be made.

That said, maybe I'll fork Gimp at some point and add some stuff that I'm missing. Or maybe I'll try to decompile XnView since that's been my sore spot for 25 years that it's not open source lol. I wonder how all this will affect small devs that make closed source apps.

Claude Code is an EXCELLENT ANALYST.
You tell it ask you questions all the time, and make suggestions. You'd be amazed how good it gets at figuring out what you want, and coming up with neat solutions. A tip is to get it to create mock-ups of the UI. Claude has become excellent at figuring out what I like, too.

I would be wary of GIMP source code, though. It's a fucking mess, you can bet your life. They should rewrite it from scratch with Claude.

 

[Otto Von Herunterhängen]

One of the supposed downsides of vibe coding is it churns out spaghetti code. It's a valid complaint, but the only reason "Spaghetti code" is bad is humans can't maintain it. When you get down to the assembler level, it's all spaghetti code, machine code is worse.

The question would be can AI maintain spaghetti code, and my bet would be the answer is yes. Which changes the whole game.

 

[TailsWin]

Claude has become excellent at figuring out what I like, too.

Idk, maybe. I've not tried CC. If I'd want to use some subscription cloud thing, I'd use something Chinese like GLM.

But anyway. The "design the UI for me" works up to a point. If you want to add some new 20 features, you're gonna need to just add more stuff, be it menu items or buttons. At some point, it'll get to a stage where:

a) you specc'd it out yourself, so you know where stuff is
b) AI put stuff where it thought is best, so you don't know where stuff is
c) AI put stuff where other similar programs put their stuff, which may not be where it makes sense for you.

Sure you can prompt AI to ask you, but regardless it's not gonna ask you where to put each of the 20 buttons.

I'm talking about the details here, not the overall concept. The first versions of Photoshop looked more like MS Paint, but the labyrinth of options that it's today is what you get by adding more details. If someone made some different decision for some detail at version 3, maybe the 'bitmap graphics suite' today would be something totally different.

I would be wary of GIMP source code, though. It's a fucking mess, you can bet your life. They should rewrite it from scratch with Claude.

Uh, why? I don't know any details, but if you want, you can just go look at their repo...

The question would be can AI maintain spaghetti code, and my bet would be the answer is yes. Which changes the whole game.

Spaghetti code is a human concept... I mean literally, spaghetti is human food lol. I find it bonkers how people blame AI for coding mistakes, when it's taught on human-written code. All the messy code, security bugs and all the mistakes from before 2 years ago are human made.

It can maintain about as well as humans can too. It's fine with a clean sheet but as stuff gets too messy, AI can get lost too, because its attention mechanisms are also trained on human data. It can't keep an entire massive repo in memory, it has to make decisions on what to focus on and how to look for what's needed.

That's one thing about AI... It would be better off if it could get to the stage where it can skip the whole 'human-readable programming language' concept and write either machine code, or some intermittent format it designed itself, with documentation format that's made for itself. But as long as its underlying tech is based on human language, it's also limited by it.

 

[TailsWin]

Ok I'll share some fun battle story snippets from my AI-assisted coding work To be clear, I'm not bashing AI here, it's just funny. These aren't even anywhere close the top models anyway.

On the other hand, experiences like this show inherent limitations of these AIs ("autoregressive transformers that generate language based on next-token prediction" if you will). Yes, you can minimise issues like this with training, and with harnesses and agentic capabilities, the model can control and check its output. But the underlying concept is still there, regardless if we're talking about 2M model, 35B (like most here) or 3T. You're getting them to generate human-readable code piece by piece, and then read it back for tweaks, based on human rules and behaviour.

Example of: "Oops I made a mistake. Eh, fuck it, maybe it'll work"




"Wait, this is supposed to be the final version?"



"What do you mean, front literally means in front of something?"

 

[FantasyForge]

Uh, why? I don't know any details, but if you want, you can just go look at their repo...

I have interacted with the GIMP guys in their forums, and it's clear they are struggling, it takes them forever to fix bugs. I get a bad impression of their developers and source code just following the disasters, they can't even fix bug 6, after over 20 years. They have all this tangled legacy spaghetti code, and struggle to upgrade. They also have an attitude problem, but so do I lol, when I point out how to fix things, and they won't.
Having said that, I'm loyal to GIMP for many kinds of image editing (but not all).

AI

As for Claude Code, it's hard to describe how good it is at complex projects: it's like sex, you need experience. Chinese models might be better at it, but I doubt that.

 

[TailsWin]

I have interacted with the GIMP guys in their forums, and it's clear they are struggling, it takes them forever to fix bugs.

Well it is a 25+ years old program which had to be updated countless times, so I wouldn't be exactly shocked if it were messy. It's also maintained by like 3 guys, I think, who don't get paid, at least not regularly (I'm not aware of any commercial project that uses Gimp as a baseline and contributes to it).

I don't know about their attitude, tho it wouldn't surprise me either. A lot of FOSS coders are arrogant dipshits. But that's why AI is so important for coding, now everyone can make their own things or forks. Isn't that what all those FOSS programmers have called for for ages? "Make your own app". Well now everyone does lol.

But anyway. It still stands that if you want a complex app like that, it will either take lots of work to make, or AI will make choices for you. Then there's all the maintenance and fixing shit. If we have gripes about the Gimp guys fixing their stuff, what chance is there for a random vibecoder properly maintaining an entire Adobe clone suite?

Until AI can truly read your mind and make apps truly special purpose for your needs on the spot, I still think it makes more sense to stick to existing solutions. Unless you want something truly special and can put in the work.

 

[Otto Von Herunterhängen]

But anyway. It still stands that if you want a complex app like that, it will either take lots of work to make, or AI will make choices for you. Then there's all the maintenance and fixing shit. If we have gripes about the Gimp guys fixing their stuff, what chance is there for a random vibecoder properly maintaining an entire Adobe clone suite?

It's not a random vibecoder doing that. There are 189 contributors to PhotoCraft at this point, FilmCraft has 75. They are all listed on the help screen for each program.

Adobe managed to piss of a lot of people. Other programs in the suite have less, but there are multiple people on each.

 

[TailsWin]

It's not a random vibecoder doing that. There are 189 contributors to PhotoCraft at this point, FilmCraft has 75. They are all listed on the help screen for each program.

We'll see. Maybe it'll work out for this particular clone, but there are lots of others that try to do the same.

Honestly though, I'm not sure if the current paradigm of FOSS with 'lead devs', 'contributors' and other roles can survive this. This may be a bigger change in software development than this whole cloning thing. Why contribute to an existing project and haggle with the devs, follow their rules etc., when you can just make your own fork and have your agent merge the updates?

If anything, having too many contributors is getting counter-productive, especially today when everyone is submitting vibecoded patches. Lots of FOSS projects are locking down their contrib guidelines to a point where only a handful of core devs really know what's what. Or I guess you can have an AI maintainer manage everything...

 

[FantasyForge]

But anyway. It still stands that if you want a complex app like that, it will either take lots of work to make, or AI will make choices for you. Then there's all the maintenance and fixing shit. If we have gripes about the Gimp guys fixing their stuff, what chance is there for a random vibecoder properly maintaining an entire Adobe clone suite?

Until AI can truly read your mind and make apps truly special purpose for your needs on the spot, I still think it makes more sense to stick to existing solutions. Unless you want something truly special and can put in the work.

Tails, we are already there. This one geek engineer used AI to create a Linux Kernel in Rust, in 11 days! A freaking Linux Kernal.
So, my money is on a random vibecoder (who knows what he's doing), and not guys battling legacy code that's a monstrous tangle of spaghetti code.

Some smartass will probably recreate "GIMP" in Rust, or C++ or whatever, I expect. It's probably already in a git repo somewhere, and ported to all platforms and without the garbage of GIMP's GTK fucking 3 or whatever it is they're stuck on.

I'm a bit salty about GIMP because of their attitude when I submitted bug reports, and they're like, "Won't' fix". But, not really, as I've always loudly promoted it over Photoshop.

 

[TailsWin]

Tails, we are already there. This one geek engineer used AI to create a Linux Kernel in Rust, in 11 days! A freaking Linux Kernal.
So, my money is on a random vibecoder (who knows what he's doing), and not guys battling legacy code that's a monstrous tangle of spaghetti code.

Some smartass will probably recreate "GIMP" in Rust, or C++ or whatever, I expect. It's probably already in a git repo somewhere, and ported to all platforms and without the garbage of GIMP's GTK fucking 3 or whatever it is they're stuck on.

I'm a bit salty about GIMP because of their attitude when I submitted bug reports, and they're like, "Won't' fix". But, not really, as I've always loudly promoted it over Photoshop.

Talking about recreating existing stuff again. What's the point, what's the added value, and what is interesting about it? Especially when we're talking about open source projects like the Linux kernel or GIMP. Translating a FOSS project into another language is monkey work. What matters is: is it useful? Does it have crippling bugs? Will it be maintained for 20-30 years like these projects?

If you have gripes with GIMP, AI rewriting it in Rust won't do shit. It needs to be polished and well maintained, otherwise you're not getting anything useful out of it. The Linux kernel runs on everything from the controller on your hard drive, to the computers running on space satellites. That's not because it's written in some particular language, it's because it's a reliable piece of engineering, and that doesn't happen magically during one Claude session.

And the GIMP guys are already talking that the next version will be a major rewrite, so they're probably doing it themselves anyway. And they know the project in and out, so I'll rather trust them than some rando. I can always fork it and add features I want, if it's something important to me.

Of all the possibilities AI can give us, people are rewriting existing projects...

 

[Otto Von Herunterhängen]

Of all the possibilities AI can give us, people are rewriting existing projects...

"Hello World"

I mean you have to start somewhere, and seeing what it can do with existing projects is a good way to do it

My bet is that the time ArtCraft gets out of the Alpha stage those projects are going to be doing things Adobe hasn't even bought out yet. Since Adobe seems utterly incapable of coming up with something new on their own.

It's a good time to be at least a little patient with the turmoil.

 

[TailsWin]

I mean you have to start somewhere, and seeing what it can do with existing projects is a good way to do it

People have been using AI for coding for years, including rewriting legacy code into newer languages. It's nothing new.

And like I said, I disagree that it's a good way to do it with programs like these. Why would it be better? There's some utility in literally cloning unique closed source projects, but even that just locks you into copying existing stuff instead of coming up with new shit. Isn't the advantage of AI so that it can do the grunt work and humans are free to do the grand creative stuff?

My bet is that the time ArtCraft gets out of the Alpha stage those projects are going to be doing things Adobe hasn't even bought out yet. Since Adobe seems utterly incapable of coming up with something new on their own.

Why do you have so much confidence in this project? I don't get it. They literally just popped up out of nowhere. They had the (business) idea and some spare money for tokens. It says nothing about their ability into the future. It's the startup-est of all startups.

Remember OpenClaw or what was the name? The most starred and cloned repo in history. A month later, nobody remembered it.

It's a good time to be at least a little patient with the turmoil.

Which is why I'm not jumping on this bandwagon and just standing by. I'm much more curious about how existing software projects adapt to all this.

 

[Otto Von Herunterhängen]

The difference between us is you are looking at this from the technology side and I am looking at it from the business model side. It's long past time to kick SAAS in the ass.

Somewhere along the lines the big tech sector got the idea to rent us our software. "You'll own nothing and be happy". Fuck that, and I mean that sincerely. I know the origin and reason for the phrase, it's a tenant of socialism, and socialism itself is topic for reasonable discussion. The problem is when capitalism takes it over. Major corporations are salivating over the idea of renting us everything, and if you can't pay the rent you are cut off. Adobe was an early adapter of the idea, but you know, streaming services of any kind are the same thing. I own CD's and I can listen to them at any time without paying Spotify or Apple Music a monthly stipend.

Now as to why I think the ArtCraft project is important: People, including me, have spent a lot of time learning things like Photoshop, and for me to go to something like Gimp or Affinity requires me to learn a whole knew way of doing things. Adobe isn't trapping me because their software is inherently better, they are trapping me because of the time it would take to learn a new system. I'd rather spend it elsewhere.

Artcraft fixes that, the parts they have already implemented are enough for me to accomplish anything I was using Photoshop for without having to learn anything new. And what Artcraft is pushing forward anybody else can too. It's not a particular project I am hot on, it's the concept that no, corporations can't leverage my skill set to rent stuff to me. I'm sure any number of people will now follow in the path of artcraft, and that's the key point.

Stuff like rewriting Linux in rust, have fun. Rewriting Windows, yeah that is long overdue, particularly since Microsoft is heavily pushing rentwear to run Windows 11. They recently dropped PC to PC Windows transfer, and tell you to use Backup instead. Which of course uses the cloud, which you only have 5GB of free space on. So fuck Microsoft too. Clone the damned thing.

Of course that begs the problem of who owns AI, and there is going to be a huge fight over that. My bet is that if we let big tech be the only avenue of renting it, the Chinese will thoroughly and completely kick our ass because it is happening already.

And as I've said before, AI is going to force fundamental changes to the world's economic system.