Serving as authoritative source seems like the only value something like this could offer. If you're doing it under section 230, you can't curate - it's just whatever people post and is of much less value.
You underestimate inertia, and the fact that the industry has trained people that they must pay continually. On top of that, the common practice is to make it sufficiently hard to get your data out of your system and into the competition's - even when there's reasonably priced competition it does a great job at adding to inertia.
The difference is without subscription, I can be pretty sure the next major version will benefit me.
With subscription,the only thing certain is that the seller wants to do as little as possible to keep taking my money. This tends to result on product updates that benefit them.
I don't know that anyone can risk running such a site. There'd need to be a long trail of proof behind each claim, and proof is getting easier to manufacture by the day.
When the dangers consist of invisible targeted manipulation of thought processes it's a whole different category of risks that kids (and most adults) are not equipped to handle. The effects are playing out around the world as we speak.
I don't know that universal tracking is the answer. I also don't think unrestricted access to children by manipulative predators (companies in this case) is the right answer. But then, I don't think they should have unrestricted access to adults either.
"Anyone can record you at any time in public" is vastly different from "a single entity is recording you over time and locations across the country/state/city"
> The easily detectable LLM writers are going to be the lazy ones.
To an extent, true. There are a lot of lazy ones though. And even for those who take steps, it sometimes leaves enough of a trace to at least raise the question.
I've realized that even when humans write that way, I also stop reading. Manipulative writing always shuts down my interest in reading it. At least when the LLM does it, it's a byproduct of training. When a human does it's intentional.
It feels manipulative. The strong language always seems like it is trying to convince you it's right even when there is no need to, such as a factual presentation. Of course if you look at popular pre-AI blogs you can see strains of the same thing, so it's no surprise that llms blend together all of those "most persuasive" methods.
The Internet is drowning in bots, everyone who hosts a site or service is paying the price. At least we have companies like this to make the problem worse.
If someone goes out of their way to hide it, it probably can't be detected. But the default commit comment and PR writeup styles are pretty distinctive.