This is an engineered system and not just some singular large fined tuned model. I like the continued progress in this direction, even the various proposed "learning agents" points in this direction. I still expect and hope it remains more likely that AGI will be a CAIS like system.