ParentFull threadsgt101·What is the evidence for 1) ? I thought that the latest models were getting "somewhere" with fairly trivial reasoning tests like ARC-1View on HN