Actually, it evolves in a random walk. Old ideas can become more practical because while we abandoned an approach, we got better at things that would make it more feasible another time around (eg VR today looks nothing like the VR of my childhood due to advances across the board in computing power, computer language expressivity, game dev tools, IMUs, IMU algorithms development, etc etc etc). That’s a case of a path opening up due to investment over time that couldn’t be accomplished by a bet.

Other times bets are insufficient in size - a bet that requires 100B to unlock trillions may seem like a failure at 10B, especially if you didn’t properly estimate the size of a market (or the market isn’t large enough for that yet).