@liaizon i think this just tests the protocol and what people want to do with it to the limits, we're seeing vibe coded AT Protocol PDSes too!
Definitely highlights rough or under defined parts of specs & tests interoperability.
@liaizon @thisismissem LLMs aside for a moment, i'm not sure i see this as a net negative. if you have a group of people who aren't programmers, whether because of where and when they were born, their genetics, their training, their economics, and make it possible for them to participate in the open source ecosystem, i think that's an amazing gift. maybe not on day one when they, like i did once upon a time, publish garbage. but eventually.
i just don't see how gatekeeping someone's ability to participate in life where and when they choose, especially in open source, is a good thing. are we really the people who should be telling them "no, you don't get to be part of this!"
@toddsundsted Is there not a parallel here to “why learn to draw if genAI can spit out pictures that are good enough,” or “why lift weights at the gym if a forklift can do 100× as much?” Results are great, but they don't have to be literally 100% of what we care about.
Lowering barriers to entry to programming is wonderful. LLMs have me considering caveats on ensuring that new approaches actually lead towards understanding and growth outside of the specific tool.
@toddsundsted As an example, I love Scratch to bits: https://scratch.mit.edu It's built and framed as an educational environment, but it's clearly capable, and its use demonstrably builds understanding of programming logic.
I believe that LLMs *can* be used in a way that leads to similar understanding and growth. I'm not sure if they are very good at it. As @liaizon notes, the approach easily tempts people into publishing things they have little to no understanding of.
@toddsundsted Heck yeah, that's how I started as well. 😄
Not everyone has to deeply know programming, but everyone is responsible for the open source code they publish. Back to @liaizon's point, there's always been shoddy open source code. But with or without LLMs, if you publish code that you can't answer people's questions on why it does what it does, that makes you a crappy open source programmer, and people would be correct to be wary about your software's reliability.