Really? AI is catching all the easy bugs I’d catch, and way more subtle ones as well. Maybe it was the case they’d miss obvious things often a year ago, but that’s not today’s frontier models.
The party speaker literally adapted to the change in eternal enemy mid sentence without breaking the grammar. How much more fast-acting and live can you get? Are you suggesting that they had a free press in 1984 that didn't have to update their own description of what was happening?
Ok maybe I went a bit over the top, but I really prefer if articles get clearly marked updates, and only update with a refresh button instead of constantly polling the server (or web socket) and changing the text I'm in the middle of reading. It's just wrong.
The reason (or, one of them) that companies pay for API rate tokens is the ZDR agreement, so Anthropic doesn’t leak company IP by way of reinforcement learning
reply