A GenAIration Lost In Space
I saw Leave Me Behind on hacker news earlier today about how a software engineer wants conversation and learning more than LLM-speed development practices.
I've lived in Intentional Community before; I think a lot about community, and I conceive of the cultural piece of DevOps mostly as a community-building exercise. The blog post made me think about different LLM-using groups and I tried to categorise them a touch:
- I don't want my community replaced with AI agents
- I see some open source communities in this bucket
- I see some individual developers in this bucket, too.
- I like that this AI helps my skills move faster
- I think of this group as one that already has a skill and uses AI to enable it further. People that rely on words for a living are very much in this group e.g. writers, software engineers, lawyers.
- I think my LLM-output means I'm a $YOUR_PROFESSION expert, too!
- These people are at best oblivious, and at worst trolling. Singing kareoke does not make you a songwriter.
- I hope this is the group that the overwhelming LLM-generated spam hitting open source software projects belongs to.
- I think my LLM-output means I'll be more successful at $YOUR_PROFESSION than you
- Some are taking advantage of current levels of interest from venture capital.
- Singers don't write songs, but they are really singing. They can and do make more money than songwriters.
I'm not trying to be exhaustive here, just capturing some real trends.
Community
The heartbreak here is watching the interaction of people in the first bucket and third bucket. It's reminiscent of how doctors couldn't get a word in on social media during the pandemic, but people selling ragebait were everywhere. What do you do when you want to be heard, but all the loudest voices are from people emulating the form of your prose and arguments, but without any of the substance?
That third group reminds me of flat-earth folks. Some are well intentioned, many are genuinely stumped by the argument 'water finds its level' because they don't see a body of water bend, but there's clearly a lot of grifters and outright liars just making a buck.
Still, watching community interaction die is heartbreaking. I'm hoping it evens out in the long run, but for now those LLMs have some loud voices.
This all reminds me of the song American Pie, by Don Mclean. He sings about changes he saw in his lifetime, a shift in the culture, a loss of innocence. The Indie-Web used to be differentiated from commercial ventures by whether you were on a small custom site or on a commercial portal. LLMs mean that our small community can't tell any more if the author wrote the words and had the ideas, unless it's egregious slop.
Me, I'm hopeful. Personal expertise is needed today. The most powerful LLMs fall apart with hallucinated nonsense pretty quickly; their best codebases hit hard security issues; their best prose sounds like absolute ass. If you don't know how to identify that 200 dependencies being pulled into a codebase to reverse a string is probably insane, or if you don't have the ear to hear the difference between slop-prose and good prose, well, I hope you learn. Understanding the details is much more fun.
A GenAIration Lost In Space
I saw Leave Me Behind on Hacker News earlier today about a software engineer who wants conversation and learning more than LLM-speed development practices.
I've lived in Intentional Community before; I think a lot about community, and I conceive of the cultural piece of DevOps mostly as a community-building exercise. The blog post made me think about different LLM-using groups and I tried to categorise them a touch:
- I don't want my community replaced with AI agents
- I see some open source communities in this bucket
- I see some individual developers in this bucket, too.
- I like that this AI helps my skills move faster
- I think of this group as one that already has a skill and uses AI to enable it further. People that rely on words for a living are very much in this group e.g. writers, software engineers, lawyers.
- I think my LLM-output means I'm a $YOUR_PROFESSION expert, too!
- These people are at best oblivious, and at worst trolling. Singing karaoke does not make you a songwriter.
- I hope this is the group that the overwhelming LLM-generated spam hitting open source software projects belongs to.
- I think my LLM-output means I'll be more successful at $YOUR_PROFESSION than you
- Some are taking advantage of current levels of interest from venture capital.
- Singers don't write songs, but they are really singing. They can and do make more money than songwriters.
I'm not trying to be exhaustive here, just capturing some real trends.
Community
The heartbreak here is watching the interaction of people in the first bucket and third bucket. It's reminiscent of how doctors couldn't get a word in on social media during the pandemic, but people selling ragebait were everywhere. What do you do when you want to be heard, but all the loudest voices are from people emulating the form of your prose and arguments, but without any of the substance?
That third group reminds me of flat-earth folks. Some are well intentioned, many are genuinely stumped by the argument 'water finds its level' because they don't see a body of water bend, but there's clearly a lot of grifters and outright liars just making a buck.
Still, watching community interaction die is heartbreaking. I'm hoping it evens out in the long run, but for now those LLMs have some loud voices.
This all reminds me of the song American Pie, by Don Mclean. He sings about changes he saw in his lifetime, a shift in the culture, a loss of innocence. The Indie-Web used to be differentiated from commercial ventures by whether you were on a small custom site or on a commercial portal. LLMs mean that our small community can't tell any more if the author wrote the words and had the ideas, unless it's egregious slop.
Me, I'm hopeful. Personal expertise is needed today. The most powerful LLMs fall apart with hallucinated nonsense pretty quickly; their best codebases hit hard security issues; their best prose sounds like absolute ass. If you don't know how to identify that 200 dependencies being pulled into a codebase to reverse a string is probably insane, or if you don't have the ear to hear the difference between slop-prose and good prose, well, I hope you learn. Understanding the details is much more fun.