I am pretty sure I know the answer, but I am wondering what people think about this. Say you come across a new fediverse platform. It has a lot of innovative features, and seems to be well thought out overall. A lot of people have been excited by the features and have been joining to check these new features out.

The one rub, the project admits that they used AI agents in the foundational stages of the project to accelerate development progress, then once the foundation was built, they analyzed, bug fixed, and documented the code, and brought in human contributors to assist with developing the project going forward. They do still allow AI code, but only certain types and using certain quality protocols to ensure code consistency, and are clear and up front about this, and unapologetic, as it was the only way they could make this project happen in the time and manner it did.

I know AI is somewhat more accepted in the coding world these days, depending on who you talk to, but I know some people have strong stances in it. How would you feel about a project like this? Under what circumstances would you be ok with joining a platform that has used AI in its code?

  • hendrik@palaver.p3x.deEnglish
    5·
    4 months ago

    I sometimes try slop-coded projects just for shits and giggles. So I’m willing to answer with a yes here. But(!) the question is very hypothetical. At least when it comes to majorly AI coded projects. Usually with these projects, I’ll soon find out half of the features are only there on paper, in a grandiose README.md written by Claude. But it’s more or less a lie and not there in reality. Or there’s some stubs and boilerplate code for it but half the features don’t work. Or the installation instructions are hallucinated, there’s no such Docker container on Docker Hub, and I stop bothering. It’ll usually go hand in hand with no one reading the issue tracker and all kinds of negative dynamics.

    So… In a hypothetical world, where AI agents code useful software with many features… Maybe? But I don’t think that’s the reality we live in.

    With projects which are mainly done by humans, and they found some way to make AI assisted programming work, I’d say maybe. Depends on the use-case. I wouldn’t risk it deleting my email inbox or trust it with important data, unless they have a really good track record. But with social-media platforms I’d be willing to risk my data. I think it’s also questionable to bet on a newer, fast moving project that gets abandoned randomly, after someone doesn’t like to pay the Anthropic bill anymore.

    For example I still use Firefox. And some more software which has AI in their workflow. I guess it’s somewhat fine in some instances. But it tends to add other unsustainable dynamics and we can see how it’s overall not healthy to companies like Mozilla. So at some point they might die anyway and I’ll have to switch to something else.