Brandon Thomas, developer of ArtCraft, a suite of open-source creative tools meant to replicate Adobe applications for free, has now announced apps to replace Microsoft Office tools such as Excel, PowerPoint, and Word, respectively, named GridCraft, DeckCraft, and WordCraft, that are created with Claude. Thomas has shared his new clones to ArtCraft GitHub, making them free for all to download.
Excuse me, do you call this open source?
For real. I can open source my valueable projects to you. But what if I deliberately make it hard to read, like mangling it or minifying it?
Hell, every piece of JavaScript code on Earth is technically open source. If you can’t read assembly, everything is open source.
Only having the source code distributed is not open source in spirit. The intention of opening the source code is welcoming auditing, learning, contributing, to name a few.
“Opening” the source code of slop, with tons of meaningless boilerplates, non-logical paths and flows, heavily coupled modules and functions, thousand-line long functions, achieves none of the purpose of the intention.
I’m guessing they believe it meets the OSI definition of open source. Essentially that the code is licensed for redistribution.
Does this have tons of meaningless boilerplate and hard to follow call chains? I have read part of the source, and what I’ve seen looks pretty well organized, but if that is how you think the current models write code, you’ve missed many months.
Perhaps that’s the problem. If it’s bad, I can still tell and choose not to read it. If it seems to make sense but actually not, I’m just going to waste my valueab time.
Edit: Just clarified a single sentence.
That you would not distribute the original source code, right? That is A source code, not THE. The generated source code by Claude Ai is THE new source code, there is no higher code generated from it. It is the implementation, the original source. It has an Open Source license. They do not mangle it and distribute a manipulated mangled version and hide the original source code.
What you criticize is the quality of code. But that is not a definition if something is Open Source. By all definitions, the distribution of original source code and a license compatible with Open Source is basically what Open Source means.
Correct. It requires a license compatible with Open Source, which it is in this case.
This is irrelevant for a definition to be Open Source or not.
There is still a distinction between bad human-written source code and AI slop.
The human holds the copyright of his work and can do whatever he wants with it, including distributing it under a open/libre license, i.e., open source.
AI slop? Not so much. The model is very likely trained on stolen work. It’s tainted. Nobody owns the copyright of everything it generates. No one can “open source” it. The license every one receives is voided.
And both who (intentionally) releases bad code to the crowd lack the spirit of open source, but the AI one is also right out illegal.
I agree with this reply here 100%. My previous reply was just about the fact that code quality isn’t important in discussion of what takes to be an Open Source code or not. Because your previous argument was “it’s bad code, how can this possibly be Open Source” (paraphrased here just).
I also think if you cannot guarantee the process does not involve in any training data from Adobe software, then how can you proof this is really clean room implementation? And besides the fact that if majority of the code is vibe coded shouldn’t be copyrightable, because machines are not humans who has copyright. And to anyone who argues that LLMs are just tools, no, they are not. It’s more complicated than that.