In the previous wave (calculators and computers), many people thought "well if AI can solve hard things like 10x10 digit multiplication and quadratic equations so much better than humans can, then surely it will very soon also do easy things like telling a cat from a dog". As it turned out, they were wrong. This time, we're seeing the same argument replay - except this time there's much stronger arguments for the "LLMs plus a few more improvements can do everything" side, we just don't know yet if those arguments are strong enough that that will turn out to be true. I feel like the utopia that "AI acceleration optimists" have in mind comes from a position of confidence that, while current wave of LLMs + agents unlock many new capabilities, there is still a significant range of capabilities that will stubbornly resist them, just like happened with the industrial revolution, and then with calculators and computers. If this is true, then the next AI era will be amazing: every foot soldi