The intellectual property of the forms and the functions of the software that are now vibe coded clones for are also copyrighted, no? - I'm thinking about GEMA vs. Suno or similar cases which imho were so far all won the royalty collecting societies. So I am skeptical.
Besides you said "Now we have the tools..." - We don't! Big tec has these tools.
Thanks for the video and the information around these cases.
While I do agree that Adobe and Microsoft can do with being taken down a notch, I don't think this business of "clean room" implementations by LLMs is the way to do it.
You mention a few times, and I agree, that we already have good open source alternatives. Krita, GIMP, LibreOffice, InkScape, to name a few. Pouring energy and resources into replicating software that is beyond enshittified will only divide us more and we will remain under the whims of the central sources whose software we're trying to replicate. Adobe still controls the PSD format, etc. In the best case, we become subjects of the LLM models, but in a worse, and more likely case, we become subjects of the owners of the compute power that runs those models, Anthropic, OpenAI, etc.
We can sit around and pretend, maybe for a few years, that we took back the power, the time will come when we lose meaningful access for one reason or another. At that point, more people will have stayed in an exploitative system than would have if these replicas didn't start popping up.
The only silver lining I see here is that companies who might be able to influence the general direction of LLM use are being affected negatively and so they might start lobbying in a somewhat favorable direction.
@thelinuxexperiment
I have embraced the AI. It gives me option to choosee. As long as the project is maintained with clear vision for future, i'm totally ok with vibe coding..
even i used AI to create extensions that i need as i couldn't find them or developer did not want to add such feature not even as an optional..
so yes AI let's go.. embracing it to the fullest!!!
Please, these tools are NOT FULLY OPEN SOURCE. They are source available, yes. But they lack the licensing conditions of free software, because the code that is used comes from the models which came from software that contain almost every license, no matter if free software, source available software or even proprietary software. Second and even more important, no human can review a codebase that is changing that fast. How can something be considered open source if humans cannot review it, because code is changing too fast? Finally I also see free software as political statement for acting ethically, but the way we use generative neuronal networks these days is everything but ethical. Even if I like the idea of taken things away from companies that took our freedom away first (I really like), it does not cost us zero dollar. In fact, we pay a lot. There are tons of such projects now, every single one requires companies to build more mega and giga data centers, all of these require a lot of energy and many projects will just die in some time and were wasting all the energy for no gain. We have to pay the bills, no matter if we use the vibecoded apps or not. Since I bought my new PC early enough, I am not affected for the next years, but I still fight for those who did not have my luck. Please don't forget the externalized costs in the math. I am not even speaking about the social costs that come on top of that.
Why not just vibecoding addons to make docx documents fully compatible to LibreOffice or adding missing features until the LibreOffice team handcraft the features? Costs much less energy than building a complete new app, is less buggy and has much more value for those who really rely on these features. Probably because it does not get so much attention and other ego-stuff. Also think about this: the projects look pretty much like "implementing low hanging fruites". The development will slow down as more spaghetti the code becomes and as more in detail the functionality has to be accurate. It is easy to create a GUI, it is easy to create basic features as selecting parts of an image, but it becomes hard when multiple tools play together for a specific outcome. Right now nobody knows if they will ever reach the same feature quality of the original applications (not speaking about code quality and resource usage, just about feature comparison from end-user perspective).