Clojurians Log v2
ai-skeptics 2026-09-10
At this point OpenAI and Anthropic would probably pay higher severance packages if their ex-employees agree to post on twitter after they're fired, spreading the word on how "scary good" their models are getting.
Most of their own marketing showcases have been like that recently. "It's so good we can't even get it to follow our rules!"
You'd think that competence in implementing guardrails for their models would do better for their image, but they've fully embraced the "scary good" narrative.
yeah, sounds like it
This particular tweet is a little suspicious IMO and it sounds like he only worked at anthropic for 6 weeks. I think its marketing/propaganda but not for the AI companies https://x.com/ParkerThayer/status/2097759699626328575?s=20
Yeah, I mean... there were multiple situations when people said "well, we gave agents no access to the internet and they found out anyway..." No, that didn't happen. This seems like a joke I heard in the past, of a bakery where bread was more expensive than bread with butter - the reason is that the baker was putting butter on the bread, then cleaning up to sell the "plain" version
About the second tweet, it's amazing how something that seems so obvious becomes political to this point.... I mean, come one, it really feels that AI became a new worldwide cult where you can't even imagine forcing them to... OH, THE HORROR!... obey the law.
