We will not see continual learning announced as fully solved in a single livestream; because labs are incentivized to release innovations quickly, we'll see a broken early version of continual learning (test time training) before something that truly learns like a human, giving lots of heads up before the bottleneck is fully solved.