Hacker News Reader

As I said in another reply, it really depends. If its a task that I have done before and new code, then I can give a fairly good estimate. If its something new, I have no idea. If it's on a monstrous old codebase, then there are a lot of unknowns to add into that.

And most people expect you to do such estimates on the spot.

More like senior engineers have learned from experience and just triple any estimate.

No one can estimate worth a damn. With experience you realize how terrible all estimates are — unless you are estimating something you have exactly done before, which is more likely with more experience. Everything I have done in my career was mostly unrelated to anything done previously, so estimating would be as effective as using dice.