Pedro Domingos

Always work first on the part of your research project that's most likely to fail. That way you don't needlessly waste time on the others.

Easier said than done... Some easy steps are usually prerequisites for the harder ones...

Sure. Preface what I said with "Subject to precedence constraints, . . ."

Yeah, but i think you need to solve the harder problem of the preference constraints before you can get to the hard to solve substantive problem you are trying to get to to begin with

What I often find is that some potential failure points are easily identified in advance, and that saves a lot of time.

More of a problem is the time you have to spend arguing with the pesky Twitter hanger on that keeps asking you dumb questions that prevent you from getting to your simple and probably correct (most of the time) point about tackling the hard problems first...

One person’s dumb questions are another’s path out of a local optimum.

If I keep asking dumb questions, will you keep jumping from local optimum to local optimum?