[thomastjeffery]

Even if that isn't 100% ubiquitously true, it's true enough to provide important context, so thanks for that.

My overall point doesn't just apply to this instance, though. It's something I see all over the place, even today; particularly in conversations about moderation and censorship.

Content has been siloed off so intensely that it's hard to even imagine a modern internet without arbitrary borders. Most of those borders are made across organizational lines. They are often made out of copyright, with the notion that some deserving party will be monetarily compensated.

Those borders usually don't align to the content itself. Instead, they become arbitrary hurdles, or even walls; making it unfeasible or impossible to truly benefit from content. Nearly every incompatibility in software was created intentionally, to cement and enforce these borders.

Now inference models (overconfidently called AI) like LLMs are all the rage. What do they do? They draw new borders. What are those borders meant to align with? The patterns that are already present in the content itself.