Dwarkesh Patel reversed his timeline for superintelligence from most of a decade to 2030, citing rapid AI progress in mathematics and the automation of frontier AI research. His essay argues that automated researchers could match the best humans by 2028.




