gradient descent into madness

It is really not all that hard to get away with murder. A basic amount of foresight and prep will get rid of enough evidence that the police either won't find the killer or will be unable to make the charge stick.

Despite this, most murders do get solved, because someone who decides that murder is the solution to their problems is probably not the kind of person to be levelheaded and careful enough to avoid getting caught.

I think something very similar is true of sockpuppet accounts. It's not that hard to make a convincing sockpuppet. However, people almost always make sockpuppets as a way of defending their egos, of making it look like they have more support than they actually do. And this passionate defense of someone the sockpuppet ostensibly doesn't know and has no reason to care about is a dead giveaway. So if you want to validate a sockpuppet you have to lose an argument to it. You have to get owned. Once it makes you look like an idiot, basically no one will ever suspect it again.

And he keeps posting in the exact same way, no matter what. Keeps bringing up the same long-debunked arguments to distract the conversation from stupid things his team says and does. Or a stereotyping personal attack/generalization.

He's not even subtle about it.

He's wasting more effort making these accounts than it takes to ignore them.

Yes, across the board

Depends on the product

No, not for any product

“These safeguards prevent foreign adversaries from using our most capable models in ways that pose severe safety risks. The US and its allies hold an edge in frontier chips and the highly optimized software that runs them at full potential,” the company said in a statement to WIRED. “These safeguards ensure Claude isn't used to erode that advantage—by optimizing chips developed by those adversaries, for example.”

and now Anthropic's noncitizen employees can't use Mythos! Which isn't to say that there is a direct causal link there or that Anthropic is responsible for the Trump administration's actions, but one might hope that it would make Anthropic warier of that kind of framing in the future.

Anthropic was just trying to get ahead of the curve on the issue - it is going to happen, best be at the table when it does so we can have a voice. It isn't like OpenAI opposes all regulation themselves either; they are opposing state-level efforts at regulation by backed federal-level regulations, and have gone in almost entirely on the "national security" angle. Because they have to! This is a done deal.