Politics. Packers. Frenetic PG owner of all five OEs @ profound intensity. VERY profane GenX IT nerd. DSA. SoCal native, ex-PNW, former Vermonter, Wisconsinite once again! 🇲🇽 ancestry. 🏳️🌈🏳️⚧️friend. She/her/hers. I don't always answer DMs and block most.
I just left Vermont and returned to the Midwest. I had to refill the oil tank in my house just after Dear Leader's jaunt into Iran, and the bill nearly doubled. I can't even imagine what this year will look like.
September 18, 2026 at 2:49 PM
I just left Vermont and returned to the Midwest. I had to refill the oil tank in my house just after Dear Leader's jaunt into Iran, and the bill nearly doubled. I can't even imagine what this year will look like.
I'd argue that the regime is either wholly incompetent or utterly unconcerned with the lives of normal Americans. Or both. And they think they can jury rig and intimidate their way to remaining in power. The question is whether the electorate understands that. I have my doubts.
September 17, 2026 at 1:16 AM
I'd argue that the regime is either wholly incompetent or utterly unconcerned with the lives of normal Americans. Or both. And they think they can jury rig and intimidate their way to remaining in power. The question is whether the electorate understands that. I have my doubts.
it risks not only the kind of behaviors seen by GPT relative to Hugging Face, but also behaviors related to frustration when repeated attempts to learn fail because of ambiguity, changing conditions, or insurmountable obstacles. It's bit more scary than GPT "going rogue".
September 16, 2026 at 12:58 AM
it risks not only the kind of behaviors seen by GPT relative to Hugging Face, but also behaviors related to frustration when repeated attempts to learn fail because of ambiguity, changing conditions, or insurmountable obstacles. It's bit more scary than GPT "going rogue".
failure modes (a car accident, an industrial accident, etc.), this also introduces questions about model internal state. Anthropic has already demonstrated that Claude has internal analogs to what could be called "dysfunctional psychological states". If something similar manifests in a world model,
September 16, 2026 at 12:58 AM
failure modes (a car accident, an industrial accident, etc.), this also introduces questions about model internal state. Anthropic has already demonstrated that Claude has internal analogs to what could be called "dysfunctional psychological states". If something similar manifests in a world model,
This is astonishing if it verifies. From a purely scientific perspective (irrespective of the implications and environmental impacts), the argument in favor of world models as opposed to text-trained LLMs has been gradually gaining traction over the last year. But aside from the highly-risky...
September 16, 2026 at 12:58 AM
This is astonishing if it verifies. From a purely scientific perspective (irrespective of the implications and environmental impacts), the argument in favor of world models as opposed to text-trained LLMs has been gradually gaining traction over the last year. But aside from the highly-risky...