You’re going to have an unacceptably high failure rate as you attempt to trial-and-error your way through all the lower-probability problems. Meanwhile, independent research paths aiming at general AI, which absolutely could handle all these problems, is racing you.