Hello fellow programs!

In April we tried out a complete ban on LLM-related content. Today we're asking for feedback on how that went, and more generally what we want to do about this kind of content. Please comment below, but if all you're going to say is "I liked/hated it", please also indicate that you've read the nuance below.

To be clear we always have and will continue to ban content that's generated by an LLM. If you don't want to write it, we don't want to read it. And we also do and will continue to ban content that's not related to programming but about e.g. philosophy in AI or jailbreaking chatgpt. (Non-programming AI articles account for most of the AI-related content that we see and we remove quite a lot of them. This is not related to the April trial.)

So the nuance is that the only additional category of content that we banned in the April trial and are asking about here is programming content that is about AI. This ranges from:

• mathematical techniques in machine learning ("using transformer techniques for sequence prediction")
• techniques for using LLMs at runtime within a small codebase
• production model deployment and testing architectures
  
• experience reports or configuration tips with Cursor
• best practises for prompting
• how we secure our AI generated codebase
• hey guise I just discovered vibe coding will AI replace programmers i am surely the first person to ask this
• how to glue an LLM to your business data
• synergisting agentic blockchains in a mobile social local world: a tedx talk featuring one line of code on the last slide

You can see that we've struggled with what to do about the various categories for a while and have moved around in our approach and we'll probably do that for a while yet. I don't want to go banning every faddy thing that's briefly so popular as to be annoying but we also need to be careful with the content that we allow because it's what drives future submitters, so it can be self feeding. This topic also brings out the rabid fans and detractors alike, so it's easy to get lost in a vocal minority. (For that reason I'm not going to pretend that this is a fully democratic decision where we add up the vote counts or something: people are too willing to brigade on this stuff and we'll keep some subjectivity to avoid that.) At some point I believe these tools will be discussed as simply as we discuss compilers or OOP or GC or VX Modules, but currently the hype and doomerism are so rabidly partisan that it's hard to find honest examples.

Note that a confounding influence is that in the last month or so the new mods really got ramped up. I was removing things like that before but on a large delay, whereas now we're better able to enforce the rules we already had. So if what you're annoyed by is "will AI replace programmers?", be aware that this has no effect on that. We already remove it.

All of that said, we want to gather ideas and feedback on how we can best handle these categories of content and suggestions for how to draw the lines so we can meet our mission to be the place with the highest quality programming content, where I can go to read something interesting and learn something new every day.