About That AI-Powered Clone of Adobe Creative Suite …

Over the last week, a project appeared a suite of applications intended to be a feature for feature replacement for the Adobe Creative Suite. All open source, all liberally licensed, cross platform, written in rust, and all created with the help of AI code tools.

They also don't work very well yet, to put it mildly. But if you ask me, this is an omen for one of the directions the software world is headed in and it raises some big questions about how disruptible the big software giants really are.

First off, the project itself. It is a collection of applications like Adobe Creative suite, available at getartcraft.com and each of which theoretically offers a functional replacement for Adobe's image and video and PDF editing programs.

I've got the most experience with Photoshop, so I looked at Photocraft, which is their Photoshop replacement.

Now, it's pretty easy to be impressed by what you see at first glance. It launches almost instantly. It looks pretty slick and functional. It opens existing PSD files without issues. Many of the controls and individual behaviors of tools look almost exactly the same.

But then you begin to run in all the weird little bugs and missing functionality that hints at how far away this thing is from being even a marginally useful product, let alone a proper replacement for Photoshop. Text adjustments behave strangely. Some operations inexplicably take forever to complete.

Some controls simply don't work at all. And so on. Now, they do say in the readme that Photocraft is in early alpha, so expect rough edges.

And one day after I first downloaded the program, another build appeared with a ton of fixes, some of which addressed the very issues I just talked about.

And a few hours after that, another version also dropped with even more fixes. That's an impressive cadence. I just don't know how sustainable it is. See, it's entirely possible. Given enough time and work, these programs could reach proper feature parity with the Adobe suite.

And that's one of the stated motives behind the project, to give people a viable alternative to paying monthly fees to Adobe for the same functionality. But as I've been hinting so far, there are many obstacles. And some of them are not about the software itself.

They're about much bigger challenges. First off, time and work are precious commodities. Outsourcing the actual work to an LLM simply means trading the limits of human time and labor for the limits of a token budget and a context window.

Someone somewhere still needs to sign off on what's happening.

And the larger the codebase gets, with more possible cross interactions and complexities with time, the harder it becomes to take responsibility for the results just by telling LLMs what to do. At some point the people behind the project have to declare a plan for its long term sustainability.

Just making it open source doesn't guarantee that how many open source projects are lying around in states of abandon because everyone involved found building a community around the project wasn't as fun as writing code?

The second big issue is something that fairly smacked me in the face when working with Photocraft. There's a big difference between something that just looks and feels like Photoshop and something that actually behaves like Photoshop, especially when many of those behaviors are subtle.

For instance, with text layers in Photoshop, one of the available options is a fake bold option.

Fake bolded text may be rendered differently by Photocraft than by Photoshop, and also differently by another free, although not open source competitor called Photopea, which I've also worked with.

So that means moving a Photoshop file between any of these programs and editing it can result in an accumulation of subtle inconsistencies. People who bounce between Microsoft Word and LibreOffice, for instance, experience this a lot too.

It's one of those subtle problems that presents itself only after it's too late. It takes human dedication to track all these subtle interactions and do something about them. Not just once, but going forward constantly.

See, I don't doubt for a minute we're going to see a lot more projects like this. Where well known commercial applications get rebuilt from the ground up with the aid of LLMs and made it to open source concerns.

The biggest targets are going to be the likes of Adobe. The companies who have long squeezed their captive audiences with subscription costs and fees and lock ins. They are ripe for being disrupted by upstart indie developers wielding LLMs.

But once you've created something like that, you have to take responsibility for it. That's where the really hard work begins. The work of making something that can sustainably replace a multi billion dollar project like Creative Suite.

Photocraft as a project needs to find a way to show that it's in for the long haul, that it's not just going to burn out in six months and force people to drop back to Creative Suite, which for all of its flaws isn't going anywhere.

LLMs showed us how code could be easy, but code was never the hard part.

What's always going to be hard is craftsmanship, attention to details, and the community building that needs to take place around big software projects to keep them vibrant and sustainable. Those are all things that still have to be done by and for people.

So to all the folks involved in the Photocraft effort. Good luck and I hope you're able to pick up where the machines leave off because you'll be the only ones to do it. Let me know what you think about this.

I'd love to hear from you in the comments. And don't forget to follow Dev with Serdar and InfoWorld on Facebook, YouTube and InfoWorld. Com.