[mironathetin]

"... I do have less ability to do brilliant-late-night-ninja-coding stunts,..."

Isn't this an advantage as well? How many ninja stunts have you seen without hundreds of bugs and being easily integrated into existing stuff and not requiring to change many existing things because they don't fit anymore with the genius code (and thus replacing well tested bits with alpha quality code)?

Being older helps to plan well before you code and get a result that does the job. As you said: getting things done.

[WalterBright]

Even long ago, I discovered that once I reached a certain level of tiredness, it was pointless to keep coding, because after a night's sleep I'd realize it was all garbage and had to revert to a previous version.

[Bartweiss]

Past that point there's no possible justification except "I need to show something functional to someone tomorrow".

I confess that, once upon a 5 A.M., I thought it would be clever to store X-Y position values not as a tuple but as two halves of a single int. There was a reason, but looking at it the next day there certainly wasn't a good enough reason. That set my low-water mark for coding while tired.

[zeveb]

Yes, this. And that goes even if one is already awake due to insomnia: one owes it to oneself as well as to one's colleagues not to write code when one's mind is not in tip-top shape.

[chao-]

>How many ninja stunts have you seen without hundreds of bugs and being easily integrated into existing stuff and not requiring to change many existing things because they don't fit anymore with the genius code (and thus replacing well tested bits with alpha quality code)?

I am familiar with the stereotype you're referencing, and have seen it from myself and others. However, to genuinely answer your question of "how many ninja stunts have you seen without hundreds of bugs..." I must answer: Many.

For the first 3 years of my company's life, I a full half of the code was done as "late night ninja coding". In between getting steady work done, I would often think about a few harder problems/features/projects, and take notes from those thoughts over a few days/weeks,on what would be the best method of accomplishing said problem/feature/project. I called it the "think twice, code once" mentality.

Eventually it would culminate (almost always in the evening) when I would either realize one last detail that made it all click, or realize that there was nothing left to plan out, and it was time to dig in and code.

Cue an overnight session where all of that planning would be brought into reality. With tests.

Even now, when my lifestyle and product allows for less of this, I still do so occasionally and it brings me great joy, significant productivity and very few bugs. Think twice, code once, and of course YMMV.

[dangravell]

Maybe that wasn't a comment on the practical benefit or otherwise of all-nighters, but on the perceived benefit from management.

If a manager doesn't think you will be up for a 36 hour shift, maybe they won't hire you.