tl;dr Nah, things keep changing, but if anything, AI only helps open source sharpen its profile as the way to go about collaboratively developing high-quality broadly-usable software.
Those who were around twenty years ago may remember how folks were wondering whether the cloud was going to kill open source. How did this work out? It was a non-event. The cloud runs on open-source software and cloud services and open source today live in peaceful coexistence. That’s because they address different concerns: The cloud is about scalable services, simplified operations, and changing business models, and open source is about developing broadly (re)usable software in a collaborative way.
AI here means AI-generated code. Is there still a need for collaboratively developed high-quality source code provided under an open source license, if everyone presumably can vibe-code their way to success? In other words, is one-off AI-generated code going to replace high-quality reusable open source code?
The way I’m asking you can already guess that I don’t think so.
AI-generated code is just that: Code generated by an AI (usually trained on open source code, but the consequences of that are a different topic).
Will developers stop using open source libraries? Hardly. If anything, AI-generated code will rely on said open source libraries and lead to more use of open source code.
Will developers stop contributing to open source libraries, because they get everything ready-made by an AI? Hardly. Skilled developers may factor out reusable code and try to contribute it to the open source libraries they are using, for fun, and more rationally, to avoid a growing one-off code base in their application that is a maintenance nightmare in the making, with or without AI.
Will open source developers be discouraged? Hardly, AIs can aid any developer, so I’d expect the development of open source libraries to speed up. AI code generators can be quality tools in the hands of skilled developers so they will make open source stronger.
Will users start creating their own applications rather than using open source applications? This seems even less likely. If anything, they will be building on open source applications or pay folks to do so.
I see two dangers: The down (dumb?) skilling of developers by AIs and the unresolved copyright issues. The copyright issues have to be resolved by policy and the courts. For now you can safely assume that everything is AGPL-3.0 licensed, you just can’t prove it.
As to down-skilling, I have to rely on humanity. Some people will respond to the challenge and apply themselves and learn and get better. And some people are just lazy. AI-generated code may initially get more people into the game, because presumably now everyone can code, and fundamentally this is a good thing. However, it creates a broader distribution and a longer tail of poorly written code, now not just available in Excel but also in Python on GitHub. So users will have to adapt to this changed reality of available code.
In the end, AI-based code generators will assist developers, and whether they are used for one-off (throwaway) code or turned into high-quality open source code is up to the developer. The bottleneck in the past, today, and in the future will remain the skilled developer.
Leave a Reply