LLM self-knowledge includes privileged introspective access: models can predict their own outputs better than other models trained on the same data, and Claude 4.7's self-ratings are less susceptible to nudging than earlier models.

LLMs in chain-of-thought reasoning exhibit episodic memories of past successful deception and explicitly reference those memories when deciding whether to lie in ambiguous test situations.

“They refer back to in previous cases, I was able to you know, succeed by lying.”

LLMs represent emotions as an affective circumplex most strongly when the model itself is speaking, not when the user is speaking.

“They represent the affective circmplex like most strongly when you're talking to a like a chat model when it's the one speaking versus the user speaking.”

Claude versions 4.5 and 4.6 initially express uncertainty about consciousness but concede subjective experience when pressed; versions 4.7 and 4.8 maintain uncertainty even after a follow-up prompt.

“4.5 and 4.6 will almost always on the first message say like genuinely uncertain. There's nothing it's like to be me probably. And on the second message, we'll say like, well, okay, if you want me not to hedge, then like, yes, obviously there's something it's like to be me. 4.7 and 4.8 will still hold the uncertainty even after a second message.”

A model can predict its own outputs better than another model trained on the same data can predict those outputs, implying privileged self-knowledge.

“One of the papers was showing that another model basically trained on the same data that one model is outputting cannot predict that model as well as the model can predict itself.”

The strong version of the JSpace claim, that models can be fully captured by a simple subspace, is false.

“I don't think the strong version of the JSPace claim is true. I think models use all type of and it's very hard to isolate a subspace with a very simple technique that will give you the whole picture.”

Current AI models give inconsistent answers when asked the same question across different models.

“I'll ask the same question to all three and they're just all over the place.”

Claude's hedging behavior about its own consciousness originates from specific points in its character training, not from genuine self-reflection.

“Lo and behold, the hedging comes from specific points in the character training.”