[jimbokun]

Nobody can perfectly predict the future.

[shimman]

So it's best to throw billions of dollars down the drain?

[mpyne]

No, it's best to manage your software delivery organization to reduce the cost and time of experiments so that you can quickly and cheaply figure out what to build... and then build that.

But there will still necessarily be things that you build that don't ship, and that's inherent to the problem domain.

[jimbokun]

Best to shelve something and not ship it once you know it won’t work, than to continue to throw even more money down the whole after it.