Quote

I think you can have 10,000 explanations for failure, but no good explanation for success.

Failures are cheap if you do them first. Failures are expensive if you do them at the end.

I'm really concerned that too-big-to-fail has become too-big-for-trial.

In a total work, the failures have their not unimportant place.

It is not enough to succeed. Others must fail.

Failure is essential. Trial and error is necessary.

Algorithms don't do a good job of detecting their own flaws.

Expect to make some mistakes when you try new and different approaches. Sometimes colossal failures lead to spectacular successes.

In certain businesses, I would say 10 failures to one success is a perfectly acceptable ratio. Because the failures die pretty quickly, they're not that expensive, and the successes can be really huge.